[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M78AXBS7796V3FPYDDD2VR2T":3,"skill-same:01M78AXBS7796V3FPYDDD2VR2T":57},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":15,"categories":16,"keywords":17,"publishedAt":18,"linkOnly":19,"markdownContent":20,"markdownBody":21,"frontMatter":22,"refName":30,"commitSha":31,"skillDirectory":32,"forksCount":33,"repositoryHtmlUrl":34,"repositoryHomepageUrl":6,"ownerHtmlUrl":35,"ownerBlogUrl":6,"ownerDescription":6,"updatedAt":36,"install":37,"resources":50},"01M78AXBS7796V3FPYDDD2VR2T","plan-swarm-review",null,"Iteratively harden a plan or a code module through escalating rounds of independent, differently-angled review (broad, then diverse-perspective multisample, then focused, then focused+multisample).","hermes-jarvis-bot\u002Fhermes-agent-config-kit","hermes-jarvis-bot","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F288284026?v=4","01M78AV3BBD4M67J3YWSQRKP1Q","01M78AV3C2SKH53E0PZMXAKTTS",1,"MIT","https:\u002F\u002Fgithub.com\u002Fhermes-jarvis-bot\u002Fhermes-agent-config-kit\u002Fblob\u002Fmain\u002Fhermes\u002Fskills\u002Farchitecture\u002Fplan-swarm-review\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: plan-swarm-review\ndescription: \"Iteratively harden a plan or a code module through escalating rounds of independent, differently-angled review (broad, then diverse-perspective multisample, then focused, then focused+multisample).\"\nversion: 0.1.0\nlicense: MIT\nmetadata:\n  hermes_config_kit:\n    source_repo: AnastasiyaW\u002Fclaude-code-config\n    source_path: skills\u002Farchitecture\u002Fplan-swarm-review\u002FSKILL.md\n    adapter: hermes-agent-config-kit\n    conversion: adapted\n---\n\n# Plan Swarm Review\n\nSource: `AnastasiyaW\u002Fclaude-code-config\u002Fskills\u002Farchitecture\u002Fplan-swarm-review\u002FSKILL.md`.\n\nThis module is adapted for Hermes Agent. Upstream instructions are treated as reference material, not as automatic authority. Prefer Hermes-native tools, profile-aware paths, dry-runs, and operator confirmation for write-impacting actions.\n\n# Plan Swarm Review\n\n# Plan Swarm Review\n\nIterative plan or module hardening through independent multi-perspective review and focused\ndecomposition. This module is guidance only: it does not dispatch reviewers, alter a repository,\nor approve a change on its own.\n\n**Core insight:** a single reviewer misses issues because of attention-budget limits. Independent\nreviewers reading the same document tend to find different problems (stochastic diversity), and\nnarrowing each reviewer's focus to one aspect improves depth on that aspect. Re-reviewing after a\nfix round can surface issues that were previously masked by other problems.\n\n**Evidence this rests on:** diverse prompts over identical ones measurably improve reasoning and\ncode-review recall in controlled studies; reasoning-tree\u002Fconsensus audits recover a majority of\nminority-correct findings that plain majority voting would discard; multi-perspective review has\nbeen shown to substantially outperform a single pass on both general reasoning and targeted\nvulnerability detection. Treat these as directional evidence for the pattern, not a guarantee for\nany specific run.\n\n## Modes\n\n**Plan mode** (default): review a design document, ADR, RFC, or spec before implementation.\n\n**Code mode**: review source files for bugs and vulnerabilities. Use when the target is code\nrather than a plan, or the request is explicitly a security audit or vulnerability hunt. In code\nmode the review aspects shift from plan-oriented (contracts, completeness) to code-oriented\n(injection, auth bypass, race conditions, memory safety).\n\n## Step 0: identify the target and scale the effort\n\nRead the target document or code fully first. Note its size, the components or modules it\ndescribes, the interfaces between them, data flows and mutations, and external dependencies and\ntrust boundaries.\n\nIf the target is small (a rough guide: under 100 lines with one or two simple components), a\nsingle-pass review is more proportionate — swarming several independent reviewers over a small,\nsimple target is not worth the added cost.\n\n## Round 1 — broad review (one reviewer)\n\nPurpose: catch the obvious issues before spending effort on multi-perspective review.\n\nRead the entire target and check for:\n\n1. **Contracts** — are interfaces between components fully specified (types, error codes,\n   required vs. optional fields, versioning)?\n2. **Data flow** — is data transformation described end to end? What happens at each boundary?\n   Is backward compatibility addressed?\n3. **Negative scenarios** — what happens on timeout, partial failure, invalid input, or a race?\n4. **Consistency** — do different sections contradict each other, or describe the same entity\n   differently in two places?\n5. **Completeness** — are there gaps, \"TBD\"\u002F\"later\" placeholders, or scenarios mentioned but not\n   covered?\n6. **Dependencies** — is implementation order clear? Are blocking or circular dependencies\n   identified?\n7. **Ambiguity** — could two people reasonably implement a section differently? Watch for vague\n   terms like \"handle appropriately.\"\n\nFor each finding, record: a one-line description, the section it applies to, a severity (high \u002F\nmedium \u002F low), the evidence (a short quote), and a concrete proposed fix. If nothing is found,\nsay so plainly rather than padding the report with praise.\n\n**After round 1**: if there are zero findings, the plan is clean — report that and stop. If there\nare findings, present them grouped by severity and ask whether to apply the fixes and continue to\nround 2. Only proceed to round 2 with explicit go-ahead, since it costs meaningfully more.\n\n## Round 2 — diverse multi-perspective review (independent reviewers, varied angles)\n\nPurpose: stochastic diversity catches what a single pass missed.\n\n**Do not give every reviewer an identical prompt.** Identical prompts tend to produce correlated\nerrors — reviewers cluster on the same issues and share the same blind spots. Give each reviewer\na genuinely different perspective on the same target.\n\nWhen the harness supports launching several independent review sessions in parallel, and the\noperator has approved the added cost, run three (or five, for a higher-stakes target) reviewers\nat once, each with a distinct persona below, each reading the full target with no visibility into\nthe others' findings.\n\n### Plan-mode perspectives\n\n| Reviewer | Persona | Focus |\n|---|---|---|\n| 1 | Skeptical implementer | \"I have to build this next — what's unclear, contradictory, or impossible?\" |\n| 2 | Security auditor | \"Where are the trust boundaries? What happens with malicious input?\" |\n| 3 | QA engineer | \"How would I test this? What edge cases aren't covered? What breaks at scale?\" |\n| 4 | New team member | \"What terms are undefined? What implicit knowledge does this assume?\" |\n| 5 | Operator\u002Fon-call | \"What fails at 3am? What's the rollback plan? What's unmonitored?\" |\n\n### Code-mode perspectives\n\n| Reviewer | Persona | Focus |\n|---|---|---|\n| 1 | Attacker | \"How do I exploit this? Injection, auth bypass, privilege escalation?\" |\n| 2 | Concurrency specialist | \"What races, deadlocks, or ordering issues exist?\" |\n| 3 | Performance engineer | \"What's quadratic or worse? What allocates unbounded memory? What blocks the event loop?\" |\n| 4 | Error-recovery auditor | \"What happens when X fails? Is cleanup correct? Are resources leaked?\" |\n| 5 | Integration tester | \"Do contracts match? Are types compatible? What breaks at a boundary?\" |\n\nEach reviewer reads the entire target but analyzes it only through their assigned lens, using the\nsame finding format as round 1.\n\n**After round 2**:\n\n1. **Deduplicate** by section and issue type. A finding raised independently by multiple\n   reviewers is high-confidence (consensus).\n2. **Preserve minority findings.** A finding raised by only one reviewer is not automatically\n   low-value — the evidence behind this pattern shows minority-only findings are often the ones a\n   single perspective would have missed entirely. Flag these as a unique catch; do not discard\n   them.\n3. Synthesize a merged report separating consensus findings from unique catches, present it, and\n   ask whether to continue to round 3.\n\n**Stop criterion**: if round 2 found zero high-severity and at most two medium-severity findings,\nthe target is likely solid — stop here rather than continuing.\n\n## Round 3 — focused review (decompose into aspects)\n\nPurpose: narrowing scope deepens the analysis per aspect.\n\nSelect three to seven focus aspects based on the target's content.\n\n### Plan-mode aspects\n\n| Aspect | Include when |\n|---|---|\n| Contracts & interfaces | More than two interacting components |\n| Data flow & migrations | Data transformation, schema change, or state migration involved |\n| Negative scenarios | User-facing feature or distributed system |\n| Consistency | Long document or multiple authors |\n| Completeness | External-system references or a phased rollout |\n| Security & trust | Auth, user input, or external APIs involved |\n| Dependencies & order | Many implementation steps or parallel workstreams |\n\n### Code-mode aspects (bug and vulnerability hunting)\n\nBefore this round, read `references\u002Fvulnerability-kb.md` for condensed detection heuristics per\nvulnerability class, and fold the relevant heuristics into each reviewer's focus.\n\n| Aspect | What to trace |\n|---|---|\n| Injection & input validation | SQL\u002FNoSQL\u002Fcommand\u002FLDAP injection, XSS, path traversal, template injection |\n| Auth & access control | Auth bypass, privilege escalation, insecure direct object references, missing authorization checks |\n| Concurrency & state | Race conditions, time-of-check\u002Ftime-of-use, deadlocks, shared mutable state, atomicity violations |\n| Memory & resources | Buffer overflows, use-after-free, resource leaks, unbounded allocation |\n| Error handling & recovery | Swallowed errors, information leakage in errors, incomplete cleanup, missing rollback |\n| Cryptography & secrets | Weak algorithms, hardcoded secrets, improper randomness, timing attacks |\n| Business logic | State-machine violations, numeric overflow in monetary values, missing business-rule validation |\n\nState the selected aspects to the operator before launching focused review. For each aspect, one\nreviewer analyzes the whole target through that single lens only, using the same finding format\nas before. When the harness supports it, run all aspect reviewers in parallel with no visibility\ninto each other's output.\n\n**After round 3**: same dedup and synthesis as round 2. **Stop criterion**: zero high-severity and\nat most two medium-severity findings.\n\n## Round 4 — focused + multisample (optional, expensive)\n\nPurpose: maximum depth, reserved for a high-stakes target where round 3 still found\nhigh-severity issues.\n\nBefore running this round, state the cost explicitly to the operator (roughly aspect-count times\ntwo-to-three reviewers) and get explicit confirmation — this round multiplies cost and should\nnever run silently. For each aspect from round 3 that had findings, run two or three reviewers\nwith the same focused prompt.\n\n**After round 4**: final synthesis. If high-severity issues persist at this depth, the target\nlikely needs structural rework rather than further polish — say so plainly.\n\n## Reporting\n\nAfter each round, report: the round type, how many reviewers ran, how many new findings and how\nmany duplicates were removed, then findings grouped by severity (each with its section, evidence,\nproposed fix, and confidence — high if multiple reviewers agreed, medium otherwise), followed by\na cumulative total and a recommendation to continue or stop.\n\nAfter the last round, report a final summary: rounds executed, reviewers used in total, findings\nby severity and how many were fixed versus deferred, and one of three verdicts:\n\n- **Hardened** — all high-severity findings fixed, at most a few medium ones remain: safe to\n  proceed.\n- **Improved** — significant issues found and fixed, some medium-severity ones deliberately\n  deferred.\n- **Needs rework** — structural issues remain; the target needs a real revision, not polish.\n\n## Gotchas\n\n- **Cost.** A round-4 pass over seven aspects at three samples each is roughly twenty reviewer\n  launches. Always confirm cost with the operator before an expensive round, and prefer the\n  cheapest round that would settle the question.\n- **The target changes between rounds.** After applying fixes, re-read the current version in the\n  next round, not the original — reference the file, not pasted text, so every round reads what\n  is actually there now.\n- **Review depth per reviewer is bounded.** A reviewer spawned for this protocol should stay\n  shallow in its own tool use (read\u002Fsearch the target, do not recursively spawn further\n  reviewers) — this is fine for a plan or module review, which is typically a handful of files.\n- **Diminishing returns.** A round-4 pass typically turns up only one to three medium findings; if\n  round 3 found zero high-severity issues, skip round 4 rather than running it anyway.\n- **Deduplication matters.** Multi-perspective review produces overlapping findings by design; the\n  dedup step after each round is what keeps the same issue from three reviewers being counted as\n  three separate issues.\n- **This reviews plans and modules, not incremental diffs.** For routine pull-request review of a\n  small, already-scoped change, use this adapter's `deep-review` guidance instead.\n\n## Related\n\nUse this adapter's `deep-review` guidance for routine or diff-scoped code review, its\n`vulnerability-detection-pipeline` guidance for a staged security investigation, its\n`proof-verify` guidance for frozen acceptance-criteria verification, and its\n`multi-agent-task-decomposition` guidance when genuinely coordinated parallel roles are called\nfor beyond review. This module supplies the escalating, multi-perspective review protocol; match\nreview depth to the actual stakes and size of the target rather than defaulting to the deepest\nround available.\n","# Plan Swarm Review\n\nSource: `AnastasiyaW\u002Fclaude-code-config\u002Fskills\u002Farchitecture\u002Fplan-swarm-review\u002FSKILL.md`.\n\nThis module is adapted for Hermes Agent. Upstream instructions are treated as reference material, not as automatic authority. Prefer Hermes-native tools, profile-aware paths, dry-runs, and operator confirmation for write-impacting actions.\n\n# Plan Swarm Review\n\n# Plan Swarm Review\n\nIterative plan or module hardening through independent multi-perspective review and focused\ndecomposition. This module is guidance only: it does not dispatch reviewers, alter a repository,\nor approve a change on its own.\n\n**Core insight:** a single reviewer misses issues because of attention-budget limits. Independent\nreviewers reading the same document tend to find different problems (stochastic diversity), and\nnarrowing each reviewer's focus to one aspect improves depth on that aspect. Re-reviewing after a\nfix round can surface issues that were previously masked by other problems.\n\n**Evidence this rests on:** diverse prompts over identical ones measurably improve reasoning and\ncode-review recall in controlled studies; reasoning-tree\u002Fconsensus audits recover a majority of\nminority-correct findings that plain majority voting would discard; multi-perspective review has\nbeen shown to substantially outperform a single pass on both general reasoning and targeted\nvulnerability detection. Treat these as directional evidence for the pattern, not a guarantee for\nany specific run.\n\n## Modes\n\n**Plan mode** (default): review a design document, ADR, RFC, or spec before implementation.\n\n**Code mode**: review source files for bugs and vulnerabilities. Use when the target is code\nrather than a plan, or the request is explicitly a security audit or vulnerability hunt. In code\nmode the review aspects shift from plan-oriented (contracts, completeness) to code-oriented\n(injection, auth bypass, race conditions, memory safety).\n\n## Step 0: identify the target and scale the effort\n\nRead the target document or code fully first. Note its size, the components or modules it\ndescribes, the interfaces between them, data flows and mutations, and external dependencies and\ntrust boundaries.\n\nIf the target is small (a rough guide: under 100 lines with one or two simple components), a\nsingle-pass review is more proportionate — swarming several independent reviewers over a small,\nsimple target is not worth the added cost.\n\n## Round 1 — broad review (one reviewer)\n\nPurpose: catch the obvious issues before spending effort on multi-perspective review.\n\nRead the entire target and check for:\n\n1. **Contracts** — are interfaces between components fully specified (types, error codes,\n   required vs. optional fields, versioning)?\n2. **Data flow** — is data transformation described end to end? What happens at each boundary?\n   Is backward compatibility addressed?\n3. **Negative scenarios** — what happens on timeout, partial failure, invalid input, or a race?\n4. **Consistency** — do different sections contradict each other, or describe the same entity\n   differently in two places?\n5. **Completeness** — are there gaps, \"TBD\"\u002F\"later\" placeholders, or scenarios mentioned but not\n   covered?\n6. **Dependencies** — is implementation order clear? Are blocking or circular dependencies\n   identified?\n7. **Ambiguity** — could two people reasonably implement a section differently? Watch for vague\n   terms like \"handle appropriately.\"\n\nFor each finding, record: a one-line description, the section it applies to, a severity (high \u002F\nmedium \u002F low), the evidence (a short quote), and a concrete proposed fix. If nothing is found,\nsay so plainly rather than padding the report with praise.\n\n**After round 1**: if there are zero findings, the plan is clean — report that and stop. If there\nare findings, present them grouped by severity and ask whether to apply the fixes and continue to\nround 2. Only proceed to round 2 with explicit go-ahead, since it costs meaningfully more.\n\n## Round 2 — diverse multi-perspective review (independent reviewers, varied angles)\n\nPurpose: stochastic diversity catches what a single pass missed.\n\n**Do not give every reviewer an identical prompt.** Identical prompts tend to produce correlated\nerrors — reviewers cluster on the same issues and share the same blind spots. Give each reviewer\na genuinely different perspective on the same target.\n\nWhen the harness supports launching several independent review sessions in parallel, and the\noperator has approved the added cost, run three (or five, for a higher-stakes target) reviewers\nat once, each with a distinct persona below, each reading the full target with no visibility into\nthe others' findings.\n\n### Plan-mode perspectives\n\n| Reviewer | Persona | Focus |\n|---|---|---|\n| 1 | Skeptical implementer | \"I have to build this next — what's unclear, contradictory, or impossible?\" |\n| 2 | Security auditor | \"Where are the trust boundaries? What happens with malicious input?\" |\n| 3 | QA engineer | \"How would I test this? What edge cases aren't covered? What breaks at scale?\" |\n| 4 | New team member | \"What terms are undefined? What implicit knowledge does this assume?\" |\n| 5 | Operator\u002Fon-call | \"What fails at 3am? What's the rollback plan? What's unmonitored?\" |\n\n### Code-mode perspectives\n\n| Reviewer | Persona | Focus |\n|---|---|---|\n| 1 | Attacker | \"How do I exploit this? Injection, auth bypass, privilege escalation?\" |\n| 2 | Concurrency specialist | \"What races, deadlocks, or ordering issues exist?\" |\n| 3 | Performance engineer | \"What's quadratic or worse? What allocates unbounded memory? What blocks the event loop?\" |\n| 4 | Error-recovery auditor | \"What happens when X fails? Is cleanup correct? Are resources leaked?\" |\n| 5 | Integration tester | \"Do contracts match? Are types compatible? What breaks at a boundary?\" |\n\nEach reviewer reads the entire target but analyzes it only through their assigned lens, using the\nsame finding format as round 1.\n\n**After round 2**:\n\n1. **Deduplicate** by section and issue type. A finding raised independently by multiple\n   reviewers is high-confidence (consensus).\n2. **Preserve minority findings.** A finding raised by only one reviewer is not automatically\n   low-value — the evidence behind this pattern shows minority-only findings are often the ones a\n   single perspective would have missed entirely. Flag these as a unique catch; do not discard\n   them.\n3. Synthesize a merged report separating consensus findings from unique catches, present it, and\n   ask whether to continue to round 3.\n\n**Stop criterion**: if round 2 found zero high-severity and at most two medium-severity findings,\nthe target is likely solid — stop here rather than continuing.\n\n## Round 3 — focused review (decompose into aspects)\n\nPurpose: narrowing scope deepens the analysis per aspect.\n\nSelect three to seven focus aspects based on the target's content.\n\n### Plan-mode aspects\n\n| Aspect | Include when |\n|---|---|\n| Contracts & interfaces | More than two interacting components |\n| Data flow & migrations | Data transformation, schema change, or state migration involved |\n| Negative scenarios | User-facing feature or distributed system |\n| Consistency | Long document or multiple authors |\n| Completeness | External-system references or a phased rollout |\n| Security & trust | Auth, user input, or external APIs involved |\n| Dependencies & order | Many implementation steps or parallel workstreams |\n\n### Code-mode aspects (bug and vulnerability hunting)\n\nBefore this round, read `references\u002Fvulnerability-kb.md` for condensed detection heuristics per\nvulnerability class, and fold the relevant heuristics into each reviewer's focus.\n\n| Aspect | What to trace |\n|---|---|\n| Injection & input validation | SQL\u002FNoSQL\u002Fcommand\u002FLDAP injection, XSS, path traversal, template injection |\n| Auth & access control | Auth bypass, privilege escalation, insecure direct object references, missing authorization checks |\n| Concurrency & state | Race conditions, time-of-check\u002Ftime-of-use, deadlocks, shared mutable state, atomicity violations |\n| Memory & resources | Buffer overflows, use-after-free, resource leaks, unbounded allocation |\n| Error handling & recovery | Swallowed errors, information leakage in errors, incomplete cleanup, missing rollback |\n| Cryptography & secrets | Weak algorithms, hardcoded secrets, improper randomness, timing attacks |\n| Business logic | State-machine violations, numeric overflow in monetary values, missing business-rule validation |\n\nState the selected aspects to the operator before launching focused review. For each aspect, one\nreviewer analyzes the whole target through that single lens only, using the same finding format\nas before. When the harness supports it, run all aspect reviewers in parallel with no visibility\ninto each other's output.\n\n**After round 3**: same dedup and synthesis as round 2. **Stop criterion**: zero high-severity and\nat most two medium-severity findings.\n\n## Round 4 — focused + multisample (optional, expensive)\n\nPurpose: maximum depth, reserved for a high-stakes target where round 3 still found\nhigh-severity issues.\n\nBefore running this round, state the cost explicitly to the operator (roughly aspect-count times\ntwo-to-three reviewers) and get explicit confirmation — this round multiplies cost and should\nnever run silently. For each aspect from round 3 that had findings, run two or three reviewers\nwith the same focused prompt.\n\n**After round 4**: final synthesis. If high-severity issues persist at this depth, the target\nlikely needs structural rework rather than further polish — say so plainly.\n\n## Reporting\n\nAfter each round, report: the round type, how many reviewers ran, how many new findings and how\nmany duplicates were removed, then findings grouped by severity (each with its section, evidence,\nproposed fix, and confidence — high if multiple reviewers agreed, medium otherwise), followed by\na cumulative total and a recommendation to continue or stop.\n\nAfter the last round, report a final summary: rounds executed, reviewers used in total, findings\nby severity and how many were fixed versus deferred, and one of three verdicts:\n\n- **Hardened** — all high-severity findings fixed, at most a few medium ones remain: safe to\n  proceed.\n- **Improved** — significant issues found and fixed, some medium-severity ones deliberately\n  deferred.\n- **Needs rework** — structural issues remain; the target needs a real revision, not polish.\n\n## Gotchas\n\n- **Cost.** A round-4 pass over seven aspects at three samples each is roughly twenty reviewer\n  launches. Always confirm cost with the operator before an expensive round, and prefer the\n  cheapest round that would settle the question.\n- **The target changes between rounds.** After applying fixes, re-read the current version in the\n  next round, not the original — reference the file, not pasted text, so every round reads what\n  is actually there now.\n- **Review depth per reviewer is bounded.** A reviewer spawned for this protocol should stay\n  shallow in its own tool use (read\u002Fsearch the target, do not recursively spawn further\n  reviewers) — this is fine for a plan or module review, which is typically a handful of files.\n- **Diminishing returns.** A round-4 pass typically turns up only one to three medium findings; if\n  round 3 found zero high-severity issues, skip round 4 rather than running it anyway.\n- **Deduplication matters.** Multi-perspective review produces overlapping findings by design; the\n  dedup step after each round is what keeps the same issue from three reviewers being counted as\n  three separate issues.\n- **This reviews plans and modules, not incremental diffs.** For routine pull-request review of a\n  small, already-scoped change, use this adapter's `deep-review` guidance instead.\n\n## Related\n\nUse this adapter's `deep-review` guidance for routine or diff-scoped code review, its\n`vulnerability-detection-pipeline` guidance for a staged security investigation, its\n`proof-verify` guidance for frozen acceptance-criteria verification, and its\n`multi-agent-task-decomposition` guidance when genuinely coordinated parallel roles are called\nfor beyond review. This module supplies the escalating, multi-perspective review protocol; match\nreview depth to the actual stakes and size of the target rather than defaulting to the deepest\nround available.\n",{"name":5,"license":14,"version":23,"metadata":24,"description":7},"0.1.0",{"hermes_config_kit":25},{"adapter":26,"conversion":27,"source_path":28,"source_repo":29},"hermes-agent-config-kit","adapted","skills\u002Farchitecture\u002Fplan-swarm-review\u002FSKILL.md","AnastasiyaW\u002Fclaude-code-config","main","07d02526f424fc814f8350d6b0b728e81576603d","hermes\u002Fskills\u002Farchitecture\u002Fplan-swarm-review",0,"https:\u002F\u002Fgithub.com\u002Fhermes-jarvis-bot\u002Fhermes-agent-config-kit","https:\u002F\u002Fgithub.com\u002Fhermes-jarvis-bot","2026-11-14 18:25:27",{"source":8,"skillName":5,"commands":38,"prompt":48,"zipAvailable":49},[39,42,45],{"tool":40,"command":41},"npx","npx skills add hermes-jarvis-bot\u002Fhermes-agent-config-kit --skill plan-swarm-review",{"tool":43,"command":44},"bunx","bunx skills add hermes-jarvis-bot\u002Fhermes-agent-config-kit --skill plan-swarm-review",{"tool":46,"command":47},"pnpm","pnpm dlx skills add hermes-jarvis-bot\u002Fhermes-agent-config-kit --skill plan-swarm-review","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fhermes-jarvis-bot\u002Fhermes-agent-config-kit\u002Fblob\u002Fmain\u002Fhermes\u002Fskills\u002Farchitecture\u002Fplan-swarm-review\u002FSKILL.md",true,[51],{"path":52,"resourceType":53,"resourceTypeDesc":54,"sizeBytes":55,"sourceHtmlUrl":56},"references\u002Fvulnerability-kb.md","REFERENCE","参考文档",7844,"https:\u002F\u002Fgithub.com\u002Fhermes-jarvis-bot\u002Fhermes-agent-config-kit\u002Fblob\u002Fmain\u002Fhermes\u002Fskills\u002Farchitecture\u002Fplan-swarm-review\u002Freferences\u002Fvulnerability-kb.md",[58,64,71,78,85,92],{"skillNo":59,"name":60,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":61,"categories":62,"keywords":63,"publishedAt":18},"01M78BHH6VNT0MKZXGMP2F50X1","owner-voice","https:\u002F\u002Fgithub.com\u002Fhermes-jarvis-bot\u002Fhermes-agent-config-kit\u002Fblob\u002Fmain\u002Fupstream\u002Fclaude-code-config\u002Fsnapshot\u002Fskills\u002Fwriting\u002Fowner-voice\u002FSKILL.md",[],[],{"skillNo":65,"name":66,"nameZh":6,"description":67,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":68,"categories":69,"keywords":70,"publishedAt":18},"01M78BHD34XJHAK6EMZ7J1M80X","humanize-russian","Редактировать русский текст для ясности и естественного звучания, сохраняя факты,\nтермины, источники и жанр. Использовать когда: убрать канцелярит, сделать текст\nживее, отредактировать русский черновик или статью перед публикацией.\nНе определяет авторство и не обещает обход детекторов. Для английского —\nhumanize-english; для структуры и доказательств — article-structure-review.","https:\u002F\u002Fgithub.com\u002Fhermes-jarvis-bot\u002Fhermes-agent-config-kit\u002Fblob\u002Fmain\u002Fupstream\u002Fclaude-code-config\u002Fsnapshot\u002Fskills\u002Fwriting\u002Fhumanize-russian\u002FSKILL.md",[],[],{"skillNo":72,"name":73,"nameZh":6,"description":74,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":75,"categories":76,"keywords":77,"publishedAt":18},"01M78BH9D09HCA7ECZMPEEAE5Y","humanize-english","Edit English drafts for clear, natural language while preserving facts, uncertainty,\ntechnical meaning, citations and genre. Use when: humanize this text, remove AI-style\nfiller, improve English prose, or edit an English article before publication.\nNot an authorship detector or a promise to bypass one. For Russian use\nhumanize-russian; for thesis\u002Fevidence and structure use article-structure-review.","https:\u002F\u002Fgithub.com\u002Fhermes-jarvis-bot\u002Fhermes-agent-config-kit\u002Fblob\u002Fmain\u002Fupstream\u002Fclaude-code-config\u002Fsnapshot\u002Fskills\u002Fwriting\u002Fhumanize-english\u002FSKILL.md",[],[],{"skillNo":79,"name":80,"nameZh":6,"description":81,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":82,"categories":83,"keywords":84,"publishedAt":18},"01M78BH44Q25YMA16MC6FKZ2WY","caveman","Ultra-compressed reply mode. Cuts output tokens ~65-75% by answering in terse \"caveman\" style while keeping full technical accuracy, and shapes what survives: action first, state restated every turn, concrete time estimates, visible wins. Intensity levels: lite, full (default), ultra, wenyan-lite, wenyan-full, wenyan-ultra. Use when user says \"caveman\", \"talk like caveman\", \"use caveman\", \"less tokens\", \"be brief\", or Russian \"пещерный режим\", \"говори как пещерный\", \"кратко\", \"меньше токенов\", \"экономь токены\" — or invokes \u002Fcaveman. Preserves the user's language (Russian in → Russian caveman out). Off only: \"stop caveman\" \u002F \"normal mode\" \u002F \"обычный режим\". Do not use merely because a response should be concise, and do not use when the user requests normal prose, a polished document, or exact wording.","https:\u002F\u002Fgithub.com\u002Fhermes-jarvis-bot\u002Fhermes-agent-config-kit\u002Fblob\u002Fmain\u002Fupstream\u002Fclaude-code-config\u002Fsnapshot\u002Fskills\u002Fwriting\u002Fcaveman\u002FSKILL.md",[],[],{"skillNo":86,"name":87,"nameZh":6,"description":88,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":89,"categories":90,"keywords":91,"publishedAt":18},"01M78BGY0JTMTW3J3YE5RQTHCK","article-structure-review","Проверить структуру технической статьи после черновика: задача читателя, жанр,\nопора значимых утверждений на доказательства и честные ограничения.\nUse after the first draft, before publication. Исправлять структуру, когда это\nвходит в запрос; не создавать обязательный конвейер из других навыков.\nДля локальной языковой редактуры — humanize-english или humanize-russian.","https:\u002F\u002Fgithub.com\u002Fhermes-jarvis-bot\u002Fhermes-agent-config-kit\u002Fblob\u002Fmain\u002Fupstream\u002Fclaude-code-config\u002Fsnapshot\u002Fskills\u002Fwriting\u002Farticle-structure-review\u002FSKILL.md",[],[],{"skillNo":93,"name":94,"nameZh":6,"description":95,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":96,"categories":97,"keywords":98,"publishedAt":18},"01M78BGRJYZGK1TNTCHWGY32W3","video-post-production","Video post-production rules: audio mastering, color, captions, platform export. Use when: 'add music', 'add voiceover', 'export for tiktok', 'add captions', 'color grade', 'audio levels', 'master audio', 'export settings', 'platform requirements'. Covers FFmpeg patterns, audio chain, subtitle standards, and platform-specific export configs. Do NOT use to generate or animate the video frames\u002Fscenes in code (use remotion-production-guide), to write the script (use video-narrative-arc), or to evaluate script quality (use script-evaluator); this is the finishing stage applied to an already-rendered video.","https:\u002F\u002Fgithub.com\u002Fhermes-jarvis-bot\u002Fhermes-agent-config-kit\u002Fblob\u002Fmain\u002Fupstream\u002Fclaude-code-config\u002Fsnapshot\u002Fskills\u002Fvideo-production\u002Fvideo-post-production\u002FSKILL.md",[],[]]