[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M7F41XT9Y25TS5ARC2MCHS95":3,"skill-same:01M7F41XT9Y25TS5ARC2MCHS95":73},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":16,"categories":17,"keywords":18,"publishedAt":19,"linkOnly":20,"markdownContent":21,"markdownBody":22,"frontMatter":23,"refName":26,"commitSha":27,"skillDirectory":28,"forksCount":14,"repositoryHtmlUrl":29,"repositoryHomepageUrl":6,"ownerHtmlUrl":30,"ownerBlogUrl":6,"ownerDescription":31,"updatedAt":32,"install":33,"resources":46},"01M7F41XT9Y25TS5ARC2MCHS95","reverse-engineer-anything",null,"Reverse engineer native, managed, Electron\u002FJavaScript, packaged, firmware, and browser targets with REA. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis.","26BB\u002Fagent-fleet-os","26BB","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F113676969?v=4","01M7F3AVX9MTTAHCYTANRGEE52","Bhushan Vijay Bhosale","01M7F3AVY15DJBX1CE4H25P52T",0,"MIT","https:\u002F\u002Fgithub.com\u002F26BB\u002Fagent-fleet-os\u002Fblob\u002Fmain\u002Fskills\u002Freverse-engineer-anything\u002FSKILL.md",[],[],"2026-11-16 10:05:34",false,"---\nname: reverse-engineer-anything\ndescription: Reverse engineer native, managed, Electron\u002FJavaScript, packaged, firmware, and browser targets with REA. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis.\nmetadata:\n  version: \"42\"\n---\n\n# REA\n\nUse REA when a claim depends on a shipped binary or package, decompilation,\npassive application runtime evidence, or comparison with\nbehavior not established by available source. For ordinary analysis of a\ncomplete source repository, use normal repository tools and do not run REA\nreadiness or provider commands.\n\n## Connect only when needed\n\nInstalling this skill supplies instructions; it does not register the REA MCP\nserver, install analysis engines, or add tools to an already-running agent.\n\nIf REA tools are available and their registration is not known to be stale,\nproceed directly to the target. Do not run diagnostics or setup before every\ninvestigation. Use the connected server's actual tool list and input schemas;\na skill installed from repository main may describe capabilities absent from\nan older npm release. Keep complete inline Evidence when that server does not\nadvertise retained references.\n\nIf tools are absent, a call is blocked by the client, or a required provider is\nunavailable, read [connection and recovery](references\u002Fconnection-and-recovery.md).\nDiagnose the observed blocker, continue useful supported work, and resume the\noriginal task after repair. Client approval, shell PATH, registration and engine\navailability have different remedies; avoid repeating setup for all of them.\n\n1. Diagnose without changing REA or client configuration. `npx` may download\n   the package and write to npm's cache. For Codex, run\n   `npx -y rea-agents@latest doctor --client codex --json`. Substitute the current\n   supported client: `claude_code`, `claude_desktop`, `codex`, `cursor`,\n   `gemini_cli`, `windsurf`, `devin`, `opencode`, `antigravity`, `copilot_cli`,\n   `commandcode`, `vscode`, `grok_build`, `omp`, or `grok_bot`. `grok_bot` has no local\n   file registration: doctor reports the Grok Bot chat step, and setup does not\n   write `mcp.json`. If the client is unknown, use `doctor --json` and\n   inspect its registration results before choosing a setup scope.\n2. Distinguish the reason. Missing, malformed, or stale registration needs a\n   scoped configuration repair. An aligned registration with no tools in the\n   active session needs a restart\u002Freconnection; doctor cannot prove that the\n   current agent has connected. A missing provider affects only tasks requiring\n   that provider: static JavaScript inspection needs neither Hopper nor Ghidra,\n   and Android inspection has separate bring-your-own JADX\u002FJava prerequisites.\n3. For a configuration repair, prepare the read-only plan:\n   `npx -y rea-agents@latest setup --client codex --dry-run --json`.\n   Use the current client's ID, show its exact proposed paths, backups, and\n   changes, and obtain approval before setup writes configuration or installs\n   Hopper. Setup normally installs the matching bundled skill too; include that\n   replacement in the reviewed plan. After approval, apply the same scope with\n   `npx -y rea-agents@latest setup --client codex --yes`. Add `--install-hopper`\n   only if that separate installation was needed and explicitly approved.\n4. Restart\u002Freconnect the affected agent. Verify that REA tools actually appear\n   in the session, then resume the original investigation. If they remain\n   absent, inspect the client's MCP launch error rather than repeating setup.\n\nREA setup never installs or upgrades Node.js, npm, Homebrew, Java, Ghidra, IDA,\nJADX, Kuna, Binwalk, or Unblob. Use existing prerequisites; do not install unrelated\nsoftware to repair\nMCP registration. For an unsupported client, use manual stdio registration\nwith a version-pinned `rea-agents` package or continue through the CLI.\n\nA concrete CLI fallback for an operator-supplied JavaScript tree or ASAR is:\n\n```bash\nnpx -y rea-agents@latest analyze-javascript-application \u002Fabsolute\u002Fpath\u002Fto\u002Fapp --json\n```\n\nNo MCP registration or native engine is required. Read the returned Evidence,\ngraph, limitations, and unknowns with the same care as an MCP result; the CLI\nreturns the Evidence record directly. This fallback does not establish that\nMCP is configured. Native CLI tasks still require their selected engine.\nPrefer an existing `rea` CLI. When none is installed and REA package acquisition\nis within the authorized scope, use `npx -y rea-agents@latest` in its place.\nConfiguration repair starts with a read-only setup plan; show paths, backups and\nskill replacement before applying authorized changes. Preserve existing approval\nand target-execution boundaries. REA setup never installs or upgrades unrelated\nsoftware, including Node.js, npm, Homebrew, Java, Ghidra, IDA, JADX, Kuna, Binwalk or Unblob.\n\n## Route the target first\n\nChoose the first tool from the target the user supplied. Use `open_binary` for\nactive-target native or archive workflows; target-free tools take their own\nexplicit path or endpoint and do not need it.\n\n- ASAR or extracted JavaScript\u002FElectron tree:\n  `analyze_javascript_application`.\n- Archive\u002Fpackage member inventory (ZIP\u002FAPK\u002FIPA\u002FMSIX\u002FAppX or DMG):\n  `open_binary` with the supplied local path; use `inspect_artifact` when its\n  graph and findings help answer the question.\n- Android APK code, classes, methods, or incoming references:\n  `inspect_android_package`, then focused Android tools when advertised.\n  Archive member inventory still uses the archive route above.\n- Managed PE\u002FCLI assembly: `inspect_managed_artifact`.\n- Firmware image: `inspect_firmware_regions` when advertised. Use\n  `extract_firmware` when extraction is requested, with a caller-selected new\n  absolute output directory. These tools use caller-supplied Binwalk\u002FUnblob on\n  Linux; see the [firmware guide](https:\u002F\u002Fgithub.com\u002Fmorluto\u002Frea\u002Fblob\u002Fmain\u002Fdocs\u002Ffirmware-analysis.md).\n- .NET NativeAOT PE\u002FELF: native Ghidra analysis. Native code can yield pseudocode;\n  consult the NativeAOT workflow in `references\u002Fnative-and-artifacts.md` for\n  optional metadata recovery and its supported host\u002Flayout boundary.\n- User-owned browser page already open: `list_browser_targets`.\n- Browser action the agent must perform (click, type, navigate):\n  `capture_browser_scenario` with declared actions. See\n  [runtime observation](references\u002Fruntime-observation.md) for connect\u002Flaunch\n  and the focused follow-up browser tools.\n- Fiddler Everywhere traffic or a native `.saz` capture: read the\n  [Fiddler workflow](references\u002Fruntime-observation.md#fiddler-everywhere-traffic-debugging).\n  Inspect saved SAZ with `inspect_web_network_capture` and `format: \"saz\"`;\n  connect Everywhere's separate MCP server for live traffic.\n- Retained HAR\u002FSAZ\u002Fnative mitmproxy\u002Faioquic qlog capture: `inspect_web_network_capture`. Select\n  original record ordinals when useful; inspect producer fields and byte\u002Fnumber\n  sidecars without fetching recorded URLs or claiming live attribution. See\n  [historical captures](https:\u002F\u002Fgithub.com\u002Fmorluto\u002Frea\u002Fblob\u002Fmain\u002Fdocs\u002Fweb-network-captures.md) for the exact upstream\n  profile and credential exclusions. Qlog support starts with aioquic 1.5.0\n  JSON 0.3; stream coverage and endpoint loss declarations do not recover bodies\n  or establish network loss. Prefer detail summary, then kind qlog trace\u002Fstream\n  pages; select one trace, stream ID and direction for contribution\u002Frange pages.\n  Facts retain all-recorded-events-before-selection scope. Other qlog layouts\n  and PCAP correlation are unsupported.\n- User-owned Electron runtime already open: `list_electron_targets`. Before an electron-main Inspector attempt, check packaged fuses with `inspect_electron_fuses`; disabled `EnableNodeCliInspectArguments` means the app ignores `--inspect`.\n- Packet-field discovery: `inspect_packet_fields` searches installed protocols\n  and field names\u002Ftypes before choosing fields for `inspect_packet_capture`.\n  Query\u002Fpage the relevant records; registrations do not establish observed traffic.\n- Recorded packet captures: `inspect_packet_capture` with selected TShark fields,\n  display filters and decode-as rules. Requires Linux, util-linux prlimit and `REA_TSHARK_COMMAND`; see\n  [capture guide](https:\u002F\u002Fgithub.com\u002Fmorluto\u002Frea\u002Fblob\u002Fmain\u002Fdocs\u002Fpacket-captures.md).\n  Packet limits count input packets before filtering. Output defaults to 50 matching\n  frames; use `match_offset` \u002F `match_limit` for later pages, or explicit null\n  limit for complete matching output. Preserve the source digest across pages.\n  This does not perform live capture.\n  For address\u002FDNS\u002FSNI\u002Fprotocol summaries use `inspect_pcap_capture` with MCP\n  detail summary, then retained kind pcap pages. Requires Wireshark 4.x tshark\n  and capinfos, with Linux prlimit or Windows Job Objects; see\n  [packet summaries](https:\u002F\u002Fgithub.com\u002Fmorluto\u002Frea\u002Fblob\u002Fmain\u002Fdocs\u002Fpcap-analysis.md).\n  Select an exact DNS question for address pages; page exhaustion does not prove\n  network completeness. This reads recordings and does not acquire traffic.\n- Decoder discovery: `inspect_signal_decoders` searches installed decoder IDs\n  or selected decoder help, including channels and options. Query and page the\n  relevant details before choosing settings; this does not scan hardware.\n- Recorded logic\u002Fsignal traces: `decode_signal_recording` with explicit decoder,\n  channel and protocol settings. On Linux with util-linux prlimit, supply sigrok-cli 0.7.2 through\n  `REA_SIGROK_COMMAND`; see [recording guide](https:\u002F\u002Fgithub.com\u002Fmorluto\u002Frea\u002Fblob\u002Fmain\u002Fdocs\u002Fsignal-recordings.md).\n  Output defaults to 50 literal lines; select annotation classes and page with\n  `line_offset` \u002F `line_limit`, or explicit null for all remaining output.\n  Window counts are display lines, not annotations or samples. Preserve source digests across pages.\n  This reads existing files; it does not acquire hardware signals.\n- Explicit local WebAssembly artifact: `inspect_wasm_artifact`. Supply WABT 1.0.42\n  through `REA_WABT_BIN_DIRECTORY`; retain exact byte\u002Ftool identity and decoded WAT.\n  Select JavaScript glue and candidate paths explicitly and preserve URL\u002Fpath\n  ambiguity. No implicit fetch or target execution; WAT is not original source.\n  See the [WASM guide](https:\u002F\u002Fgithub.com\u002Fmorluto\u002Frea\u002Fblob\u002Fmain\u002Fdocs\u002Fwasm-artifacts.md).\n- Explicit local EVM bytecode carrier: `inspect_evm_interface` with caller-selected\n  `raw` or `hex` encoding. Preserve carrier\u002Fdecoded digests and treat selectors,\n  argument strings and mutability as inferred candidates; this performs no chain\n  lookup or target execution. For the complete inferred interface without hex or\n  duplicate stdout, discover support and select `detail: \"summary\"`; complete\n  Evidence stays retained under its original identity. Use the returned\n  `parent_evidence_id` for exact function or `bytecode_hex`\u002Fstdout\u002Fstderr pages.\n  Candidate coverage does not establish interface completeness. See the\n  [EVM bytecode guide](https:\u002F\u002Fgithub.com\u002Fmorluto\u002Frea\u002Fblob\u002Fmain\u002Fdocs\u002Fevm-bytecode.md).\n- Embedded Go compiler, module or build-setting metadata: `inspect_go_binary`\n  with the explicit local file. This built-in target-free inspection requires no\n  Go installation or disassembler. Preserve raw module bytes and source ranges;\n  absent metadata does not prove a binary is not Go, and unparsed text records\n  remain explicit. See the\n  [Go binary guide](https:\u002F\u002Fgithub.com\u002Fmorluto\u002Frea\u002Fblob\u002Fmain\u002Fdocs\u002Fgo-binaries.md).\n- Embedded cargo-auditable Rust dependency metadata: `inspect_rust_binary`\n  with the explicit local ELF, PE, or thin Mach-O file. This built-in\n  target-free inspection requires no Rust installation or disassembler. Preserve\n  package order, dependency indices, and the section file range. A missing\n  `.dep-v0` section is null metadata, not proof the binary has no dependencies.\n  A rustc commit hash is an inference from a read-only `\u002Frustc\u002F\u003C40-hex>\u002F` path.\n  See the\n  [Rust binary guide](https:\u002F\u002Fgithub.com\u002Fmorluto\u002Frea\u002Fblob\u002Fmain\u002Fdocs\u002Frust-binaries.md).\n\n- Explicit local SQLite database snapshot: `inspect_sqlite_database`. Inspect\n  with Node.js 24.x >=24.15 or 26+ providing native SQLite limits. Read schema\n  first (compact inventory by default), or select an exact ordinary table\n  and `columns` when rows are needed. Defaults are selected-table schema and\n  20 rows; request `schema_scope: \"database\"` for the complete stored schema. Preserve both database and existing sibling WAL identity; row samples\n  report truncation and have unspecified order. Use a quiescent snapshot rather\n  than assuming a live database\u002FWAL pair was copied atomically. See the\n  [SQLite guide](https:\u002F\u002Fgithub.com\u002Fmorluto\u002Frea\u002Fblob\u002Fmain\u002Fdocs\u002Fsqlite-databases.md).\n- Explicit Linux ELF file for offline layout, symbols, relocations or static\n  mitigation evidence: `inspect_binary_layout`. This target-free operation uses\n  caller-supplied pwntools without opening a disassembler database. Preserve its\n  raw locations and inference\u002Fcoverage limits; see the\n  [offline binary guide](https:\u002F\u002Fgithub.com\u002Fmorluto\u002Frea\u002Fblob\u002Fmain\u002Fdocs\u002Fbinary-diagnostics.md).\n- Supplied Linux x86-64 ELF core: `inspect_recorded_crash`. Read every recorded\n  thread, raw note and source-bound register without opening a live target.\n  Historical PIDs do not select live processes. Request `include_debugger_context`\n  only when core-only mapping candidates help; see the\n  [recorded crash guide](https:\u002F\u002Fgithub.com\u002Fmorluto\u002Frea\u002Fblob\u002Fmain\u002Fdocs\u002Frecorded-crashes.md).\n- PE resource type\u002Fname\u002Flanguage trees and static icon group references:\n  `inspect_pe_resources` with an explicit path. This portable operation requires\n  no active target or engine; opaque payloads retain file ranges and SHA-256.\n- Native executable, library, or analysis database: `open_binary`, then\n  use focused analysis tools directly; call `binary_overview` when metadata or\n  inventory context is useful and available from the selected provider.\n\nFor an existing IDA MCP configuration, use `open_binary` with the original\ninput binary and `provider_id: \"ida\"`; do not pass an `.idb` or `.i64` database.\nThe legacy attached profile binds the already-open GUI input by SHA-256 and\nleaves its database open. The modern headless profile creates and closes an\nowned private database without saving. Installation and registration reuse\n[mrexodia\u002Fida-pro-mcp](https:\u002F\u002Fgithub.com\u002Fmrexodia\u002Fida-pro-mcp); see the\n[IDA provider guide](https:\u002F\u002Fgithub.com\u002Fmorluto\u002Frea\u002Fblob\u002Fmain\u002Fdocs\u002Fida-provider.md).\nProceed directly to `analyze_function`, pseudocode, or function\u002Fstring searches\nafter opening; IDA does not supply `binary_overview`. Consult session availability\nwhen composing broader workflows. Modern direct callers and unsupported dossier\nfacets remain unknown; live IDA results are not replayed from snapshots.\n\nWith `REA_KUNA_COMMAND` configured, `provider_id: \"kuna\"` answers procedure\nlists, lookups, pseudocode, assembly and `analyze_function` from Kuna's cached\nwhole-program export. Callers\u002Fcallees use a separate body-free graph shared by\nthe session; `xrefs` returns exact-address incoming sites and filters forwarding\naliases. These queries and inventory lookups do not trigger a C export. Indirect\ntargets remain unknown and can make relationship coverage partial. Reference\ncaches expire on close; a fresh CLI process repeats the native reference walk.\nTyped references, CFG, raw-memory tools and annotations remain unsupported. A cold first\nanswer can come from a single-function run (its limitations say so); assembly\nwaits for the export. For pseudocode and assembly, `raw_result.source` names the\ncached `.c`\u002F`.asm` file and byte span, which you may search directly instead of\ncalling per function. See the\n[Kuna provider guide](https:\u002F\u002Fgithub.com\u002Fmorluto\u002Frea\u002Fblob\u002Fmain\u002Fdocs\u002Fkuna-provider.md).\n\nIf the user says only \"my app\" without a name, path, or target already selected\nfor this task, ask which app before searching for artifacts. Paths in unrelated\nsaved reports and nearby test fixtures do not select the user's target. Resolve\nan actual human-readable app name to one clear installed artifact when possible;\nask when matches are ambiguous. Never choose an example app on the user's behalf.\n\nIn a target-free session, use `open_binary` to bind any archive\u002Fpackage or\nnative target whose analysis tool operates on the active target. Do not call an\nunadvertised tool; inspect `binary_session` with `{}` and its\n`result.tool_availability` for availability reasons and remediation. Expand each\ngroup's `names` into entries, inheriting the inventory-level client requirements\nunless the group overrides them. In its closed effect and annotation vocabularies,\nlisted flags are true and omitted known flags are false. The grouped name union\nis the complete tool inventory; use `tools\u002Flist` for canonical tool schemas.\nProvider candidate facts describe possible providers, while top-level session\ncapabilities describe the current route and `analysis_provider_binding` names the\nselected provider. Do not treat those scopes as interchangeable.\n\n## Work summary-first\n\nStart with the default result and use its inline Evidence and graph context.\nFor a JavaScript application or ELF layout whose size is unknown or large,\ncall `analyze_javascript_application` or `inspect_binary_layout` with\n`\"detail\": \"summary\"`, then inspect the retained `parent_evidence_id` with\n`inspect_analysis_view`.\nPage module identities and select the returned `node_id` before requesting an\nitem; a single file path can represent several graph roles. Do not repeat an\nidentical analysis or rejected call without a changed input or resolved blocker.\nRecheck session state when a lifecycle change makes that status useful.\nMake a focused follow-up only when the\nreturned result leaves a specific question unanswered.\n\nIf MCP reports `resource_constraint` with `details.resource: \"transport\"`, call\n`inspect_analysis_view` with its reported same-session Evidence reference for a\nsummary, one section\u002Fmodule, or a stable page. Use `trace_application_feature`\nonce a seed is known, or call `export_evidence_bundle` to write the complete\nsession to a selected path. The analysis remains complete in the ledger; avoid\nrepeating it merely to request the same oversized response. Complete CLI JSON\noutput also streams without a single MCP frame.\n\nWhen the retained parent ID is unknown, use `get_evidence_bundle` with\n`detail: \"summary\"` and exact filters for operation, target SHA-256,\nanalysis-profile digest or canonical procedure address. Select its returned\nEvidence ID for the next view instead of requesting the complete bundle.\n`retention: \"complete-record\"` means the whole record is stored; native\nvalue-flow truncation and unknown metadata remain separate. For an exported\nportable bundle, use `rea inspect-evidence-bundle` with JSON containing `path`,\n`detail: \"summary\"` and the same `filters`. Reading that file still requires\nvalidating its full contents within the local memory budget.\n\nA client can also truncate a successful tool result without a REA transport\nerror. Treat that preview as incomplete; use the retained parent ID for smaller\nviews or export when the full record is needed. Do not parse the preview as\ncanonical JSON or repeat the producer. For JavaScript, read the\n[module-view and trace recipes](references\u002Fjavascript-applications.md) before\ndrilling into a large graph; a broad bidirectional trace can expand unrelated\ncontainment relationships even after summary analysis.\n\nEvery conclusion must distinguish observations, inferences, and unknowns. Cite\nEvidence IDs, preserve limitations and incomplete coverage, and never imply\nthat static analysis observed execution. Runtime requests execute the declared\ntarget and lifecycle; do not broaden the target or action beyond those fields.\n\nWithin the user's requested investigation, call available analysis tools\ndirectly. REA does not require permission grants or per-call approval flags.\nThe agent client can independently require approval. Follow the declared request\nscope and the host's actual access requirements; use the recovery guide for an\nobserved client-policy rejection.\n\n## Plan broader investigations\n\nFor requests that span multiple features or subsystems, use a staged workflow:\n\n1. Turn the request into a checklist of questions and the evidence each answer\n   needs. Resolve target identity and constraints from the conversation and\n   workspace before asking for information again.\n2. Inspect the current REA session, artifact identity, saved analysis database,\n   bookmarks, and prior evidence. Reuse matching state; do not open duplicate\n   sessions or repeat identical analysis.\n3. Start with the smallest useful overview or inventory. Follow each question\n   from its entry point through relevant data and state changes to its result.\n   Batch related operations around a specific hypothesis, then expand only when\n   the returned evidence leaves a concrete gap.\n4. Inspect relevant packaged resources and configuration alongside code when\n   they affect the question. Use format-aware inventory and parsers; do not\n   infer behavior from filenames, strings, or layout alone.\n5. Corroborate a conclusion with the evidence type it requires. Use runtime\n   observation only when static evidence cannot answer the question and the\n   required host runtime and OS access are available. For JavaScript behavior,\n   run probes against the actual app through the available browser, Electron,\n   or process capture workflow.\n6. Decompose work into independent questions. When parallel workers are\n   available and the questions do not depend on one another, assign distinct\n   scopes, point workers to existing evidence, and ask them to return sources,\n   conclusions, and unresolved gaps. Otherwise, work sequentially.\n7. Keep a concise finding ledger linking each conclusion to Evidence IDs,\n   confidence\u002Fevidence type, search boundary, and remaining unknowns. Update\n   the shared index or investigation report so later passes can reuse results.\n\nBefore finishing, revisit the original checklist. Mark each question as\nanswered, partially answered, or unresolved based on its evidence; keep\nbounded negative searches bounded, and do not describe a broad investigation\nas complete while required questions remain open.\n\n## Read only the relevant guide\n\n- Native binaries, managed assemblies, archives, and extraction:\n  [references\u002Fnative-and-artifacts.md](references\u002Fnative-and-artifacts.md)\n- ASARs, extracted JavaScript, feature tracing, and version comparison:\n  [references\u002Fjavascript-applications.md](references\u002Fjavascript-applications.md)\n- Android APK declarations, classes, methods, and static references:\n  [references\u002Fandroid-applications.md](references\u002Fandroid-applications.md)\n- Browser\u002FElectron observation (passive and scenario) and static\u002Fruntime\n  reconciliation:\n  [references\u002Fruntime-observation.md](references\u002Fruntime-observation.md)\n- Evidence paging, comparisons, residual unknowns, and verification:\n  [references\u002Fevidence-workflows.md](references\u002Fevidence-workflows.md)\n\n## Finish the task\n\nExplain findings in plain language and tie them to returned evidence. When the\nuser asks to build something, use normal coding tools and separate observed\nbehavior from design choices. Close an opened native session with\n`close_binary` when the investigation is complete.\n","# REA\n\nUse REA when a claim depends on a shipped binary or package, decompilation,\npassive application runtime evidence, or comparison with\nbehavior not established by available source. For ordinary analysis of a\ncomplete source repository, use normal repository tools and do not run REA\nreadiness or provider commands.\n\n## Connect only when needed\n\nInstalling this skill supplies instructions; it does not register the REA MCP\nserver, install analysis engines, or add tools to an already-running agent.\n\nIf REA tools are available and their registration is not known to be stale,\nproceed directly to the target. Do not run diagnostics or setup before every\ninvestigation. Use the connected server's actual tool list and input schemas;\na skill installed from repository main may describe capabilities absent from\nan older npm release. Keep complete inline Evidence when that server does not\nadvertise retained references.\n\nIf tools are absent, a call is blocked by the client, or a required provider is\nunavailable, read [connection and recovery](references\u002Fconnection-and-recovery.md).\nDiagnose the observed blocker, continue useful supported work, and resume the\noriginal task after repair. Client approval, shell PATH, registration and engine\navailability have different remedies; avoid repeating setup for all of them.\n\n1. Diagnose without changing REA or client configuration. `npx` may download\n   the package and write to npm's cache. For Codex, run\n   `npx -y rea-agents@latest doctor --client codex --json`. Substitute the current\n   supported client: `claude_code`, `claude_desktop`, `codex`, `cursor`,\n   `gemini_cli`, `windsurf`, `devin`, `opencode`, `antigravity`, `copilot_cli`,\n   `commandcode`, `vscode`, `grok_build`, `omp`, or `grok_bot`. `grok_bot` has no local\n   file registration: doctor reports the Grok Bot chat step, and setup does not\n   write `mcp.json`. If the client is unknown, use `doctor --json` and\n   inspect its registration results before choosing a setup scope.\n2. Distinguish the reason. Missing, malformed, or stale registration needs a\n   scoped configuration repair. An aligned registration with no tools in the\n   active session needs a restart\u002Freconnection; doctor cannot prove that the\n   current agent has connected. A missing provider affects only tasks requiring\n   that provider: static JavaScript inspection needs neither Hopper nor Ghidra,\n   and Android inspection has separate bring-your-own JADX\u002FJava prerequisites.\n3. For a configuration repair, prepare the read-only plan:\n   `npx -y rea-agents@latest setup --client codex --dry-run --json`.\n   Use the current client's ID, show its exact proposed paths, backups, and\n   changes, and obtain approval before setup writes configuration or installs\n   Hopper. Setup normally installs the matching bundled skill too; include that\n   replacement in the reviewed plan. After approval, apply the same scope with\n   `npx -y rea-agents@latest setup --client codex --yes`. Add `--install-hopper`\n   only if that separate installation was needed and explicitly approved.\n4. Restart\u002Freconnect the affected agent. Verify that REA tools actually appear\n   in the session, then resume the original investigation. If they remain\n   absent, inspect the client's MCP launch error rather than repeating setup.\n\nREA setup never installs or upgrades Node.js, npm, Homebrew, Java, Ghidra, IDA,\nJADX, Kuna, Binwalk, or Unblob. Use existing prerequisites; do not install unrelated\nsoftware to repair\nMCP registration. For an unsupported client, use manual stdio registration\nwith a version-pinned `rea-agents` package or continue through the CLI.\n\nA concrete CLI fallback for an operator-supplied JavaScript tree or ASAR is:\n\n```bash\nnpx -y rea-agents@latest analyze-javascript-application \u002Fabsolute\u002Fpath\u002Fto\u002Fapp --json\n```\n\nNo MCP registration or native engine is required. Read the returned Evidence,\ngraph, limitations, and unknowns with the same care as an MCP result; the CLI\nreturns the Evidence record directly. This fallback does not establish that\nMCP is configured. Native CLI tasks still require their selected engine.\nPrefer an existing `rea` CLI. When none is installed and REA package acquisition\nis within the authorized scope, use `npx -y rea-agents@latest` in its place.\nConfiguration repair starts with a read-only setup plan; show paths, backups and\nskill replacement before applying authorized changes. Preserve existing approval\nand target-execution boundaries. REA setup never installs or upgrades unrelated\nsoftware, including Node.js, npm, Homebrew, Java, Ghidra, IDA, JADX, Kuna, Binwalk or Unblob.\n\n## Route the target first\n\nChoose the first tool from the target the user supplied. Use `open_binary` for\nactive-target native or archive workflows; target-free tools take their own\nexplicit path or endpoint and do not need it.\n\n- ASAR or extracted JavaScript\u002FElectron tree:\n  `analyze_javascript_application`.\n- Archive\u002Fpackage member inventory (ZIP\u002FAPK\u002FIPA\u002FMSIX\u002FAppX or DMG):\n  `open_binary` with the supplied local path; use `inspect_artifact` when its\n  graph and findings help answer the question.\n- Android APK code, classes, methods, or incoming references:\n  `inspect_android_package`, then focused Android tools when advertised.\n  Archive member inventory still uses the archive route above.\n- Managed PE\u002FCLI assembly: `inspect_managed_artifact`.\n- Firmware image: `inspect_firmware_regions` when advertised. Use\n  `extract_firmware` when extraction is requested, with a caller-selected new\n  absolute output directory. These tools use caller-supplied Binwalk\u002FUnblob on\n  Linux; see the [firmware guide](https:\u002F\u002Fgithub.com\u002Fmorluto\u002Frea\u002Fblob\u002Fmain\u002Fdocs\u002Ffirmware-analysis.md).\n- .NET NativeAOT PE\u002FELF: native Ghidra analysis. Native code can yield pseudocode;\n  consult the NativeAOT workflow in `references\u002Fnative-and-artifacts.md` for\n  optional metadata recovery and its supported host\u002Flayout boundary.\n- User-owned browser page already open: `list_browser_targets`.\n- Browser action the agent must perform (click, type, navigate):\n  `capture_browser_scenario` with declared actions. See\n  [runtime observation](references\u002Fruntime-observation.md) for connect\u002Flaunch\n  and the focused follow-up browser tools.\n- Fiddler Everywhere traffic or a native `.saz` capture: read the\n  [Fiddler workflow](references\u002Fruntime-observation.md#fiddler-everywhere-traffic-debugging).\n  Inspect saved SAZ with `inspect_web_network_capture` and `format: \"saz\"`;\n  connect Everywhere's separate MCP server for live traffic.\n- Retained HAR\u002FSAZ\u002Fnative mitmproxy\u002Faioquic qlog capture: `inspect_web_network_capture`. Select\n  original record ordinals when useful; inspect producer fields and byte\u002Fnumber\n  sidecars without fetching recorded URLs or claiming live attribution. See\n  [historical captures](https:\u002F\u002Fgithub.com\u002Fmorluto\u002Frea\u002Fblob\u002Fmain\u002Fdocs\u002Fweb-network-captures.md) for the exact upstream\n  profile and credential exclusions. Qlog support starts with aioquic 1.5.0\n  JSON 0.3; stream coverage and endpoint loss declarations do not recover bodies\n  or establish network loss. Prefer detail summary, then kind qlog trace\u002Fstream\n  pages; select one trace, stream ID and direction for contribution\u002Frange pages.\n  Facts retain all-recorded-events-before-selection scope. Other qlog layouts\n  and PCAP correlation are unsupported.\n- User-owned Electron runtime already open: `list_electron_targets`. Before an electron-main Inspector attempt, check packaged fuses with `inspect_electron_fuses`; disabled `EnableNodeCliInspectArguments` means the app ignores `--inspect`.\n- Packet-field discovery: `inspect_packet_fields` searches installed protocols\n  and field names\u002Ftypes before choosing fields for `inspect_packet_capture`.\n  Query\u002Fpage the relevant records; registrations do not establish observed traffic.\n- Recorded packet captures: `inspect_packet_capture` with selected TShark fields,\n  display filters and decode-as rules. Requires Linux, util-linux prlimit and `REA_TSHARK_COMMAND`; see\n  [capture guide](https:\u002F\u002Fgithub.com\u002Fmorluto\u002Frea\u002Fblob\u002Fmain\u002Fdocs\u002Fpacket-captures.md).\n  Packet limits count input packets before filtering. Output defaults to 50 matching\n  frames; use `match_offset` \u002F `match_limit` for later pages, or explicit null\n  limit for complete matching output. Preserve the source digest across pages.\n  This does not perform live capture.\n  For address\u002FDNS\u002FSNI\u002Fprotocol summaries use `inspect_pcap_capture` with MCP\n  detail summary, then retained kind pcap pages. Requires Wireshark 4.x tshark\n  and capinfos, with Linux prlimit or Windows Job Objects; see\n  [packet summaries](https:\u002F\u002Fgithub.com\u002Fmorluto\u002Frea\u002Fblob\u002Fmain\u002Fdocs\u002Fpcap-analysis.md).\n  Select an exact DNS question for address pages; page exhaustion does not prove\n  network completeness. This reads recordings and does not acquire traffic.\n- Decoder discovery: `inspect_signal_decoders` searches installed decoder IDs\n  or selected decoder help, including channels and options. Query and page the\n  relevant details before choosing settings; this does not scan hardware.\n- Recorded logic\u002Fsignal traces: `decode_signal_recording` with explicit decoder,\n  channel and protocol settings. On Linux with util-linux prlimit, supply sigrok-cli 0.7.2 through\n  `REA_SIGROK_COMMAND`; see [recording guide](https:\u002F\u002Fgithub.com\u002Fmorluto\u002Frea\u002Fblob\u002Fmain\u002Fdocs\u002Fsignal-recordings.md).\n  Output defaults to 50 literal lines; select annotation classes and page with\n  `line_offset` \u002F `line_limit`, or explicit null for all remaining output.\n  Window counts are display lines, not annotations or samples. Preserve source digests across pages.\n  This reads existing files; it does not acquire hardware signals.\n- Explicit local WebAssembly artifact: `inspect_wasm_artifact`. Supply WABT 1.0.42\n  through `REA_WABT_BIN_DIRECTORY`; retain exact byte\u002Ftool identity and decoded WAT.\n  Select JavaScript glue and candidate paths explicitly and preserve URL\u002Fpath\n  ambiguity. No implicit fetch or target execution; WAT is not original source.\n  See the [WASM guide](https:\u002F\u002Fgithub.com\u002Fmorluto\u002Frea\u002Fblob\u002Fmain\u002Fdocs\u002Fwasm-artifacts.md).\n- Explicit local EVM bytecode carrier: `inspect_evm_interface` with caller-selected\n  `raw` or `hex` encoding. Preserve carrier\u002Fdecoded digests and treat selectors,\n  argument strings and mutability as inferred candidates; this performs no chain\n  lookup or target execution. For the complete inferred interface without hex or\n  duplicate stdout, discover support and select `detail: \"summary\"`; complete\n  Evidence stays retained under its original identity. Use the returned\n  `parent_evidence_id` for exact function or `bytecode_hex`\u002Fstdout\u002Fstderr pages.\n  Candidate coverage does not establish interface completeness. See the\n  [EVM bytecode guide](https:\u002F\u002Fgithub.com\u002Fmorluto\u002Frea\u002Fblob\u002Fmain\u002Fdocs\u002Fevm-bytecode.md).\n- Embedded Go compiler, module or build-setting metadata: `inspect_go_binary`\n  with the explicit local file. This built-in target-free inspection requires no\n  Go installation or disassembler. Preserve raw module bytes and source ranges;\n  absent metadata does not prove a binary is not Go, and unparsed text records\n  remain explicit. See the\n  [Go binary guide](https:\u002F\u002Fgithub.com\u002Fmorluto\u002Frea\u002Fblob\u002Fmain\u002Fdocs\u002Fgo-binaries.md).\n- Embedded cargo-auditable Rust dependency metadata: `inspect_rust_binary`\n  with the explicit local ELF, PE, or thin Mach-O file. This built-in\n  target-free inspection requires no Rust installation or disassembler. Preserve\n  package order, dependency indices, and the section file range. A missing\n  `.dep-v0` section is null metadata, not proof the binary has no dependencies.\n  A rustc commit hash is an inference from a read-only `\u002Frustc\u002F\u003C40-hex>\u002F` path.\n  See the\n  [Rust binary guide](https:\u002F\u002Fgithub.com\u002Fmorluto\u002Frea\u002Fblob\u002Fmain\u002Fdocs\u002Frust-binaries.md).\n\n- Explicit local SQLite database snapshot: `inspect_sqlite_database`. Inspect\n  with Node.js 24.x >=24.15 or 26+ providing native SQLite limits. Read schema\n  first (compact inventory by default), or select an exact ordinary table\n  and `columns` when rows are needed. Defaults are selected-table schema and\n  20 rows; request `schema_scope: \"database\"` for the complete stored schema. Preserve both database and existing sibling WAL identity; row samples\n  report truncation and have unspecified order. Use a quiescent snapshot rather\n  than assuming a live database\u002FWAL pair was copied atomically. See the\n  [SQLite guide](https:\u002F\u002Fgithub.com\u002Fmorluto\u002Frea\u002Fblob\u002Fmain\u002Fdocs\u002Fsqlite-databases.md).\n- Explicit Linux ELF file for offline layout, symbols, relocations or static\n  mitigation evidence: `inspect_binary_layout`. This target-free operation uses\n  caller-supplied pwntools without opening a disassembler database. Preserve its\n  raw locations and inference\u002Fcoverage limits; see the\n  [offline binary guide](https:\u002F\u002Fgithub.com\u002Fmorluto\u002Frea\u002Fblob\u002Fmain\u002Fdocs\u002Fbinary-diagnostics.md).\n- Supplied Linux x86-64 ELF core: `inspect_recorded_crash`. Read every recorded\n  thread, raw note and source-bound register without opening a live target.\n  Historical PIDs do not select live processes. Request `include_debugger_context`\n  only when core-only mapping candidates help; see the\n  [recorded crash guide](https:\u002F\u002Fgithub.com\u002Fmorluto\u002Frea\u002Fblob\u002Fmain\u002Fdocs\u002Frecorded-crashes.md).\n- PE resource type\u002Fname\u002Flanguage trees and static icon group references:\n  `inspect_pe_resources` with an explicit path. This portable operation requires\n  no active target or engine; opaque payloads retain file ranges and SHA-256.\n- Native executable, library, or analysis database: `open_binary`, then\n  use focused analysis tools directly; call `binary_overview` when metadata or\n  inventory context is useful and available from the selected provider.\n\nFor an existing IDA MCP configuration, use `open_binary` with the original\ninput binary and `provider_id: \"ida\"`; do not pass an `.idb` or `.i64` database.\nThe legacy attached profile binds the already-open GUI input by SHA-256 and\nleaves its database open. The modern headless profile creates and closes an\nowned private database without saving. Installation and registration reuse\n[mrexodia\u002Fida-pro-mcp](https:\u002F\u002Fgithub.com\u002Fmrexodia\u002Fida-pro-mcp); see the\n[IDA provider guide](https:\u002F\u002Fgithub.com\u002Fmorluto\u002Frea\u002Fblob\u002Fmain\u002Fdocs\u002Fida-provider.md).\nProceed directly to `analyze_function`, pseudocode, or function\u002Fstring searches\nafter opening; IDA does not supply `binary_overview`. Consult session availability\nwhen composing broader workflows. Modern direct callers and unsupported dossier\nfacets remain unknown; live IDA results are not replayed from snapshots.\n\nWith `REA_KUNA_COMMAND` configured, `provider_id: \"kuna\"` answers procedure\nlists, lookups, pseudocode, assembly and `analyze_function` from Kuna's cached\nwhole-program export. Callers\u002Fcallees use a separate body-free graph shared by\nthe session; `xrefs` returns exact-address incoming sites and filters forwarding\naliases. These queries and inventory lookups do not trigger a C export. Indirect\ntargets remain unknown and can make relationship coverage partial. Reference\ncaches expire on close; a fresh CLI process repeats the native reference walk.\nTyped references, CFG, raw-memory tools and annotations remain unsupported. A cold first\nanswer can come from a single-function run (its limitations say so); assembly\nwaits for the export. For pseudocode and assembly, `raw_result.source` names the\ncached `.c`\u002F`.asm` file and byte span, which you may search directly instead of\ncalling per function. See the\n[Kuna provider guide](https:\u002F\u002Fgithub.com\u002Fmorluto\u002Frea\u002Fblob\u002Fmain\u002Fdocs\u002Fkuna-provider.md).\n\nIf the user says only \"my app\" without a name, path, or target already selected\nfor this task, ask which app before searching for artifacts. Paths in unrelated\nsaved reports and nearby test fixtures do not select the user's target. Resolve\nan actual human-readable app name to one clear installed artifact when possible;\nask when matches are ambiguous. Never choose an example app on the user's behalf.\n\nIn a target-free session, use `open_binary` to bind any archive\u002Fpackage or\nnative target whose analysis tool operates on the active target. Do not call an\nunadvertised tool; inspect `binary_session` with `{}` and its\n`result.tool_availability` for availability reasons and remediation. Expand each\ngroup's `names` into entries, inheriting the inventory-level client requirements\nunless the group overrides them. In its closed effect and annotation vocabularies,\nlisted flags are true and omitted known flags are false. The grouped name union\nis the complete tool inventory; use `tools\u002Flist` for canonical tool schemas.\nProvider candidate facts describe possible providers, while top-level session\ncapabilities describe the current route and `analysis_provider_binding` names the\nselected provider. Do not treat those scopes as interchangeable.\n\n## Work summary-first\n\nStart with the default result and use its inline Evidence and graph context.\nFor a JavaScript application or ELF layout whose size is unknown or large,\ncall `analyze_javascript_application` or `inspect_binary_layout` with\n`\"detail\": \"summary\"`, then inspect the retained `parent_evidence_id` with\n`inspect_analysis_view`.\nPage module identities and select the returned `node_id` before requesting an\nitem; a single file path can represent several graph roles. Do not repeat an\nidentical analysis or rejected call without a changed input or resolved blocker.\nRecheck session state when a lifecycle change makes that status useful.\nMake a focused follow-up only when the\nreturned result leaves a specific question unanswered.\n\nIf MCP reports `resource_constraint` with `details.resource: \"transport\"`, call\n`inspect_analysis_view` with its reported same-session Evidence reference for a\nsummary, one section\u002Fmodule, or a stable page. Use `trace_application_feature`\nonce a seed is known, or call `export_evidence_bundle` to write the complete\nsession to a selected path. The analysis remains complete in the ledger; avoid\nrepeating it merely to request the same oversized response. Complete CLI JSON\noutput also streams without a single MCP frame.\n\nWhen the retained parent ID is unknown, use `get_evidence_bundle` with\n`detail: \"summary\"` and exact filters for operation, target SHA-256,\nanalysis-profile digest or canonical procedure address. Select its returned\nEvidence ID for the next view instead of requesting the complete bundle.\n`retention: \"complete-record\"` means the whole record is stored; native\nvalue-flow truncation and unknown metadata remain separate. For an exported\nportable bundle, use `rea inspect-evidence-bundle` with JSON containing `path`,\n`detail: \"summary\"` and the same `filters`. Reading that file still requires\nvalidating its full contents within the local memory budget.\n\nA client can also truncate a successful tool result without a REA transport\nerror. Treat that preview as incomplete; use the retained parent ID for smaller\nviews or export when the full record is needed. Do not parse the preview as\ncanonical JSON or repeat the producer. For JavaScript, read the\n[module-view and trace recipes](references\u002Fjavascript-applications.md) before\ndrilling into a large graph; a broad bidirectional trace can expand unrelated\ncontainment relationships even after summary analysis.\n\nEvery conclusion must distinguish observations, inferences, and unknowns. Cite\nEvidence IDs, preserve limitations and incomplete coverage, and never imply\nthat static analysis observed execution. Runtime requests execute the declared\ntarget and lifecycle; do not broaden the target or action beyond those fields.\n\nWithin the user's requested investigation, call available analysis tools\ndirectly. REA does not require permission grants or per-call approval flags.\nThe agent client can independently require approval. Follow the declared request\nscope and the host's actual access requirements; use the recovery guide for an\nobserved client-policy rejection.\n\n## Plan broader investigations\n\nFor requests that span multiple features or subsystems, use a staged workflow:\n\n1. Turn the request into a checklist of questions and the evidence each answer\n   needs. Resolve target identity and constraints from the conversation and\n   workspace before asking for information again.\n2. Inspect the current REA session, artifact identity, saved analysis database,\n   bookmarks, and prior evidence. Reuse matching state; do not open duplicate\n   sessions or repeat identical analysis.\n3. Start with the smallest useful overview or inventory. Follow each question\n   from its entry point through relevant data and state changes to its result.\n   Batch related operations around a specific hypothesis, then expand only when\n   the returned evidence leaves a concrete gap.\n4. Inspect relevant packaged resources and configuration alongside code when\n   they affect the question. Use format-aware inventory and parsers; do not\n   infer behavior from filenames, strings, or layout alone.\n5. Corroborate a conclusion with the evidence type it requires. Use runtime\n   observation only when static evidence cannot answer the question and the\n   required host runtime and OS access are available. For JavaScript behavior,\n   run probes against the actual app through the available browser, Electron,\n   or process capture workflow.\n6. Decompose work into independent questions. When parallel workers are\n   available and the questions do not depend on one another, assign distinct\n   scopes, point workers to existing evidence, and ask them to return sources,\n   conclusions, and unresolved gaps. Otherwise, work sequentially.\n7. Keep a concise finding ledger linking each conclusion to Evidence IDs,\n   confidence\u002Fevidence type, search boundary, and remaining unknowns. Update\n   the shared index or investigation report so later passes can reuse results.\n\nBefore finishing, revisit the original checklist. Mark each question as\nanswered, partially answered, or unresolved based on its evidence; keep\nbounded negative searches bounded, and do not describe a broad investigation\nas complete while required questions remain open.\n\n## Read only the relevant guide\n\n- Native binaries, managed assemblies, archives, and extraction:\n  [references\u002Fnative-and-artifacts.md](references\u002Fnative-and-artifacts.md)\n- ASARs, extracted JavaScript, feature tracing, and version comparison:\n  [references\u002Fjavascript-applications.md](references\u002Fjavascript-applications.md)\n- Android APK declarations, classes, methods, and static references:\n  [references\u002Fandroid-applications.md](references\u002Fandroid-applications.md)\n- Browser\u002FElectron observation (passive and scenario) and static\u002Fruntime\n  reconciliation:\n  [references\u002Fruntime-observation.md](references\u002Fruntime-observation.md)\n- Evidence paging, comparisons, residual unknowns, and verification:\n  [references\u002Fevidence-workflows.md](references\u002Fevidence-workflows.md)\n\n## Finish the task\n\nExplain findings in plain language and tie them to returned evidence. When the\nuser asks to build something, use normal coding tools and separate observed\nbehavior from design choices. Close an opened native session with\n`close_binary` when the investigation is complete.\n",{"name":5,"metadata":24,"description":7},{"version":25},"42","main","98985c16dc8df1bbd7dcd1bc698cd997a093e751","skills\u002Freverse-engineer-anything","https:\u002F\u002Fgithub.com\u002F26BB\u002Fagent-fleet-os","https:\u002F\u002Fgithub.com\u002F26BB","🚀 Aspiring Product Manager | CS @ GS Moze\r\n\r\n🛠️ Building 0→1 products across SaaS & clean-tech\r\n\r\n📊 Turning user feedback & data into shipped features","2026-11-05 13:34:38",{"source":8,"skillName":5,"commands":34,"prompt":44,"zipAvailable":45},[35,38,41],{"tool":36,"command":37},"npx","npx skills add 26BB\u002Fagent-fleet-os --skill reverse-engineer-anything",{"tool":39,"command":40},"bunx","bunx skills add 26BB\u002Fagent-fleet-os --skill reverse-engineer-anything",{"tool":42,"command":43},"pnpm","pnpm dlx skills add 26BB\u002Fagent-fleet-os --skill reverse-engineer-anything","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002F26BB\u002Fagent-fleet-os\u002Fblob\u002Fmain\u002Fskills\u002Freverse-engineer-anything\u002FSKILL.md",true,[47,53,57,61,65,69],{"path":48,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":51,"sourceHtmlUrl":52},"references\u002Fandroid-applications.md","REFERENCE","参考文档",2125,"https:\u002F\u002Fgithub.com\u002F26BB\u002Fagent-fleet-os\u002Fblob\u002Fmain\u002Fskills\u002Freverse-engineer-anything\u002Freferences\u002Fandroid-applications.md",{"path":54,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":55,"sourceHtmlUrl":56},"references\u002Fconnection-and-recovery.md",8823,"https:\u002F\u002Fgithub.com\u002F26BB\u002Fagent-fleet-os\u002Fblob\u002Fmain\u002Fskills\u002Freverse-engineer-anything\u002Freferences\u002Fconnection-and-recovery.md",{"path":58,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":59,"sourceHtmlUrl":60},"references\u002Fevidence-workflows.md",4363,"https:\u002F\u002Fgithub.com\u002F26BB\u002Fagent-fleet-os\u002Fblob\u002Fmain\u002Fskills\u002Freverse-engineer-anything\u002Freferences\u002Fevidence-workflows.md",{"path":62,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":63,"sourceHtmlUrl":64},"references\u002Fjavascript-applications.md",11962,"https:\u002F\u002Fgithub.com\u002F26BB\u002Fagent-fleet-os\u002Fblob\u002Fmain\u002Fskills\u002Freverse-engineer-anything\u002Freferences\u002Fjavascript-applications.md",{"path":66,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":67,"sourceHtmlUrl":68},"references\u002Fnative-and-artifacts.md",18911,"https:\u002F\u002Fgithub.com\u002F26BB\u002Fagent-fleet-os\u002Fblob\u002Fmain\u002Fskills\u002Freverse-engineer-anything\u002Freferences\u002Fnative-and-artifacts.md",{"path":70,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":71,"sourceHtmlUrl":72},"references\u002Fruntime-observation.md",4798,"https:\u002F\u002Fgithub.com\u002F26BB\u002Fagent-fleet-os\u002Fblob\u002Fmain\u002Fskills\u002Freverse-engineer-anything\u002Freferences\u002Fruntime-observation.md",[74,81,88,95,102,109],{"skillNo":75,"name":76,"nameZh":6,"description":77,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":78,"categories":79,"keywords":80,"publishedAt":19},"01M7F51NQSCRM2PWFMATCV9MEG","taste-design","Semantic Design System Skill for Google Stitch. Generates agent-friendly DESIGN.md files that enforce premium, anti-generic UI standards — strict typography, calibrated color, asymmetric layouts, perpetual micro-motion, and hardware-accelerated performance.","https:\u002F\u002Fgithub.com\u002F26BB\u002Fagent-fleet-os\u002Fblob\u002Fmain\u002Fstitch agent\u002Fplugins\u002Fstitch-utilities\u002Fplaybooks\u002Ftaste-design\u002FSKILL.md",[],[],{"skillNo":82,"name":83,"nameZh":6,"description":84,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":85,"categories":86,"keywords":87,"publishedAt":19},"01M7F51JGTS9VJ3S9T6VMQWZ08","stitch-loop","Teaches agents to iteratively build websites using Stitch with an autonomous baton-passing loop pattern","https:\u002F\u002Fgithub.com\u002F26BB\u002Fagent-fleet-os\u002Fblob\u002Fmain\u002Fstitch agent\u002Fplugins\u002Fstitch-utilities\u002Fplaybooks\u002Fstitch-loop\u002FSKILL.md",[],[],{"skillNo":89,"name":90,"nameZh":6,"description":91,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":92,"categories":93,"keywords":94,"publishedAt":19},"01M7F51EZR2XYDWMA3XKWH1HX2","site-md","Analyze a project description or site requirements and synthesize a project constitution into SITE.md for the Stitch Build Loop","https:\u002F\u002Fgithub.com\u002F26BB\u002Fagent-fleet-os\u002Fblob\u002Fmain\u002Fstitch agent\u002Fplugins\u002Fstitch-utilities\u002Fplaybooks\u002Fsite-md\u002FSKILL.md",[],[],{"skillNo":96,"name":97,"nameZh":6,"description":98,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":99,"categories":100,"keywords":101,"publishedAt":19},"01M7F51BNCMP0GQ7SGP39SZYDP","enhance-prompt","Transforms vague UI ideas into polished, Stitch-optimized prompts. Enhances specificity, adds UI\u002FUX keywords, injects design system context, and structures output for better generation results.","https:\u002F\u002Fgithub.com\u002F26BB\u002Fagent-fleet-os\u002Fblob\u002Fmain\u002Fstitch agent\u002Fplugins\u002Fstitch-utilities\u002Fplaybooks\u002Fenhance-prompt\u002FSKILL.md",[],[],{"skillNo":103,"name":104,"nameZh":6,"description":105,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":106,"categories":107,"keywords":108,"publishedAt":19},"01M7F518SX707EB3X42B2V7HRA","design-md","Analyze Stitch projects and synthesize a semantic design system into DESIGN.md files","https:\u002F\u002Fgithub.com\u002F26BB\u002Fagent-fleet-os\u002Fblob\u002Fmain\u002Fstitch agent\u002Fplugins\u002Fstitch-utilities\u002Fplaybooks\u002Fdesign-md\u002FSKILL.md",[],[],{"skillNo":110,"name":111,"nameZh":6,"description":112,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":113,"categories":114,"keywords":115,"publishedAt":19},"01M7F5148PVYTA76NR3X0YK8SK","stitch::upload-to-stitch","Upload local assets (images, mockups, extracted HTML, design markdown) to a Stitch project. ALWAYS use this skill when you need to upload visual assets, HTML pages, or design docs to Stitch, particularly when direct MCP tool calls fail or truncate due to base64 token limits.","https:\u002F\u002Fgithub.com\u002F26BB\u002Fagent-fleet-os\u002Fblob\u002Fmain\u002Fstitch agent\u002Fplugins\u002Fstitch-design\u002Fplaybooks\u002Fupload-to-stitch\u002FSKILL.md",[],[]]