[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M72E5VTCK4YCDV0N0B2G4HB3":3,"skill-same:01M72E5VTCK4YCDV0N0B2G4HB3":49},{"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":25,"commitSha":26,"skillDirectory":27,"forksCount":28,"repositoryHtmlUrl":29,"repositoryHomepageUrl":30,"ownerHtmlUrl":31,"ownerBlogUrl":32,"ownerDescription":33,"updatedAt":34,"install":35,"resources":48},"01M72E5VTCK4YCDV0N0B2G4HB3","refresh-backends",null,"Assess and update agentgrep's backends and storage catalogue against the agent clients installed on this machine (claude, codex, cursor-agent, gemini, agy, grok, pi, opencode, VS Code, Cursor IDE, Antigravity IDE, Windsurf), versioning what each client version stores. Use when asked to sweep, refresh, audit, or update the backends or storages, check for drift after a client upgrade, observe a new client version, or record which version a storage change landed in.","tony\u002Fagentgrep","tony","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F26336?v=4","01M72E5RNJ7SQV8F9KBM5RKRMN","Tony Narlock","01M72E5RP6M0TRDFA54GT6B85H",9,"MIT","https:\u002F\u002Fgithub.com\u002Ftony\u002Fagentgrep\u002Fblob\u002Fmaster\u002F.agents\u002Fskills\u002Frefresh-backends\u002FSKILL.md",[],[],"2026-11-10 18:39:20",false,"---\nname: refresh-backends\ndescription: >-\n  Assess and update agentgrep's backends and storage catalogue against the\n  agent clients installed on this machine (claude, codex, cursor-agent,\n  gemini, agy, grok, pi, opencode, VS Code, Cursor IDE, Antigravity IDE,\n  Windsurf), versioning what each client version stores. Use when asked to\n  sweep, refresh, audit, or update the backends or storages, check for drift\n  after a client upgrade, observe a new client version, or record which\n  version a storage change landed in.\nargument-hint: \"[agent ...]\"\n---\n\n# Refreshing agent backends\n\nEach installed client version gets a schema-only manifest in\n`docs\u002F_observations\u002F\u003Cagent>\u002F\u003Cversion>.toml`, and the catalogue stamps in\n`src\u002Fagentgrep\u002Fstore_catalog\u002F` must match the newest one. A refresh records\nnew versions, catalogues what changed, ties each change to the version that\nfirst showed it, and proves search still works and is no slower.\n\n## Arguments\n\n`$ARGUMENTS` lists agent ids as `uv run scripts\u002Fobserve_stores.py --help`\nprints them; `--agent` takes one id, so run each command once per id. Empty\nmeans `--agent all`. \"In scope\" below means these agents.\n\n## Rules\n\n- Agent homes, `\u002Fmnt\u002Fc` included, are read-only: read files and run\n  `--version`, never another client subcommand. No `HOME=` overrides.\n- No prompt text, PII, or real home paths anywhere: manifests use `${HOME}`\n  tokens, docs pages keep their existing `~\u002F` form.\n- Never hand-edit a manifest, and never rewrite or delete one for a version\n  that is no longer installed. One that already leaks a path is the user's\n  call to scrub: report it.\n- Never push, tag, or open a PR. No backup branches.\n- If the user asked for `CHANGES`, its entries land last, in their own\n  `docs(CHANGES) \u003Cdescription>` commit under the existing unreleased\n  heading, with no version lead paragraph. Otherwise leave `CHANGES` alone\n  and list the entries it needs in the report.\n\n## Commits\n\nCommit in this order, skipping a kind with nothing in it. Subjects stay at\n50 characters or fewer, and bodies take the `why:`\u002F`what:` form from\n`.github\u002FWRITING.md`.\n\n1. `agentgrep(fix[observe]): ...` when the observer itself misbehaves,\n   with a test in `tests\u002Ftest_storage_observations.py`.\n2. `agentgrep(chore[catalog]): Refresh observed stamps`: every new\n   manifest, restamp, and version line. The body names each agent and\n   version recorded, and says earlier manifests stay untouched.\n3. `agentgrep(feat[catalog]): Catalogue \u003Cn> new stores`: new rows, the\n   `catalog_version` bump, and the touched agents re-observed at their\n   unchanged versions.\n4. `agentgrep(feat[\u003Cagent>]): ...` or `agentgrep(fix[\u003Cagent>]): ...`:\n   adapter changes with their fixture and test.\n5. `docs(backends): Correct claims the sweep disproved`.\n6. `docs(backends): Record storage changes by version`.\n7. `docs(backends): Observe \u003CAgent> \u003Cversion>` for a client that updates\n   mid-sweep: its manifest, restamp, version lines, and entry in one commit.\n8. `docs(CHANGES) \u003Cdescription>`, when the user asked for it.\n\n## Steps\n\n1. **Worktree.** Unless the user names a branch, branch from fresh\n   `origin\u002Fmaster` into a sibling worktree, with the session date as\n   `YYYYMMDD`. Work only there.\n\n   ```console\n   $ git fetch origin\n   ```\n\n   ```console\n   $ git worktree add --no-track -b storages-YYYYMMDD ..\u002Fagentgrep-storages-YYYYMMDD origin\u002Fmaster\n   ```\n\n2. **Baseline.** Exit 1 means drift. Table each in-scope agent's documented\n   version (newest manifest; `_\u003CAGENT>_OBSERVED_VERSION` in\n   `src\u002Fagentgrep\u002Fstore_catalog\u002F\u003Cagent>.py`, hyphens as underscores) against\n   the installed version this prints.\n\n   ```console\n   $ uv run scripts\u002Fobserve_stores.py check --agent \u003Cid>\n   ```\n\n   `windsurf` has no version probe, so `check` never reports its upgrades.\n   Read `windsurfVersion` from the newest\n   `~\u002F.windsurf-server\u002Fbin\u002F\u003Ccommit>\u002Fproduct.json`, and decide whether that\n   build wrote its catalogued data before stamping it. The first `check`\n   after the IDE probes landed reports each IDE once as drift from its\n   `unknown.toml`.\n\n3. **Observe** each in-scope agent whose version changed or that `check`\n   reports drifting. A new version writes a new manifest; drift at the same\n   version rewrites that version's file. A refusal to write `unknown.toml`\n   over a versioned manifest means the version probe failed: report it,\n   never work around it.\n\n   ```console\n   $ uv run scripts\u002Fobserve_stores.py observe --agent \u003Cid>\n   ```\n\n   A new version diffs against the previous manifest:\n\n   ```console\n   $ git diff --no-index docs\u002F_observations\u002F\u003Cid>\u002F\u003Cold>.toml docs\u002F_observations\u002F\u003Cid>\u002F\u003Cnew>.toml\n   ```\n\n   A same-version rewrite diffs against the commit:\n\n   ```console\n   $ git diff -- docs\u002F_observations\u002F\u003Cid>\u002F\u003Cversion>.toml\n   ```\n\n4. **Triage** every diff line:\n\n   - Before calling a key added or removed, check older manifests. The\n     observer samples each store's newest 40 files, so keys drift in and out\n     of the sample without a schema change.\n   - Version changed, shape didn't: restamp only.\n   - New unclaimed path: add a `catalog_only` row, or `private` when it\n     holds credentials or keys. A new home directory for the agent also goes\n     in its `AgentProbe.homes` in `scripts\u002Fobserve_stores.py`.\n   - Changed discriminator, keys, tables, or columns in a store agentgrep\n     reads: update the descriptor and adapter, add a redacted fixture under\n     `tests\u002Fsamples\u002F\u003Cagent>\u002F\u003Cstore_id>\u002F`, and show the new test failing\n     before it passes.\n   - `source_count \u003Cold> -> \u003Cnew>` on a store agentgrep reads means its\n     files moved, were renamed, or were recompressed, and search lost them.\n     Make discovery and the adapter follow them, or, when that needs a\n     design, list it first under what was left undone.\n   - New readable conversation store: give it the `primary_chat` or\n     `supplementary_chat` role, never `prompt_history`, so default search\n     does not open it.\n   - A file newer than its newest record is an empty draft session, not a\n     parse failure.\n   - Re-read the agent's backend page and each touched row's\n     `schema_notes`, and correct any claim the observation disproves.\n\n   Any descriptor add, removal, or reshape follows \"Adding or updating a\n   store\" in `docs\u002Fdev\u002Fstorage-catalog.md`, including its `upstream_ref`\n   (pinned to an upstream commit, `github.com\u002F\u003Corg>\u002F\u003Crepo>@\u003Csha>\u002F...`) or\n   `sample_record`. Bump `catalog_version` and re-observe the agent in that\n   same commit. If the unclaimed list now names the children of a directory,\n   the new row claims that directory by name without covering its children:\n   put a `\u003Cplaceholder>` or glob at that level of its pattern.\n\n5. **Restamp** only what you re-read:\n\n   - `_\u003CAGENT>_OBSERVED_VERSION`, keeping its prefix (`claude-code v...`).\n   - `_\u003CAGENT>_OBSERVED_AT`, `_CATALOG_CAPTURED_AT`, and the date in the\n     comment above the dated block, in\n     `src\u002Fagentgrep\u002Fstore_catalog\u002F_common.py`.\n   - The `observed_version` line wherever `docs\u002Fbackends\u002F\u003Cagent>.md` and\n     `docs\u002Fdev\u002Fstorage-catalog.md` carry one. Windsurf's page is\n     `docs\u002Fbackends\u002Funsupported\u002Fwindsurf.md`.\n\n   An agent that is not installed, or whose version you could not read,\n   keeps its version and date: move its `_OBSERVED_AT` out of the dated\n   block with a comment saying why, as Windsurf's is. When the gate fails on\n   a row that keeps its own `observed_at`, re-verify that row on disk before\n   pointing it at the module constant.\n\n6. **Version the change** under `## Changes by version` on the agent's page,\n   adding that section, with the opening paragraph other pages use, where it\n   is missing. Each `### \u003Cnew>` entry opens \"Seen in \u003Cnew> (YYYY-MM-DD);\n   absent in \u003Cold> (YYYY-MM-DD).\" Versions between two observations were\n   never seen, so a change landed after `\u003Cold>` and by `\u003Cnew>`; never name an\n   exact version. With no readable version, the entry is `### Observed\n   YYYY-MM-DD`, bracketed \"seen \u003Cdate>, absent \u003Cdate>\", naming any build\n   version you could read. A version with no storage change still gets an\n   entry saying so. A store agentgrep does not read is \"not searched\", never\n   \"excluded\".\n\n7. **Prove search** for every touched agent id except `windsurf`, which has\n   no search surface; check it through its docs page. Pick `\u003Cterm>` from a\n   store you changed, and keep it out of commits and the report.\n\n   ```console\n   $ uv run agentgrep find --agent \u003Cid> --json\n   ```\n\n   ```console\n   $ uv run agentgrep search --agent \u003Cid> --scope all --exhaustive --json \u003Cterm>\n   ```\n\n   Pass means records returned, `summary.coverage.sources_failed` is 0, and\n   `summary.diagnostics` has nothing new. For a field you started reading\n   (model, cwd, title), returned records must carry real values; a key in\n   the output proves nothing. Then run the Layer 0 MCP smoke from the\n   `testing-mcp-with-cli-agents` skill: a throwaway FastMCP stdio client in\n   the scratchpad doing `tools\u002Flist` plus one `search`.\n\n8. **Gate, then commit.** Before every commit: `rm -rf docs\u002F_build`, the\n   gates in `.github\u002FCONTRIBUTING.md`, `just test-all`, and\n   `just build-docs`. When `FORCE_COLOR` is set, run tests under\n   `NO_COLOR=1`. Read each command's own pass\u002Ffail line, never a `| tail` or\n   `| grep` of it. `just build-docs` lines starting `storage observations:`\n   mean a stamp and its newest manifest disagree. Then stage and scan the\n   added lines; it must print nothing. It catches encoded project\n   directories such as `home-\u003Cuser>-work-...` and `%2Fhome%2F...`.\n\n   ```console\n   $ git diff --cached | grep -nE \"^\\+.*(\u002Fhome\u002F|\u002Fmnt\u002F|\u002FUsers\u002F|home-$USER-|%2F(home|Users)%2F)\"\n   ```\n\n   Before the last commit, `check` exits 0 for every in-scope agent; a\n   client that updated mid-sweep gets commit 7.\n\n9. **Measure** after the last commit, with every configured bench and its\n   configured runs and warmup, which each bench's timeout is sized for; the\n   `benchmark` skill covers flags and `analyze`. Default `search` must not\n   slow beyond run-to-run noise; explain any remaining delta with\n   `scripts\u002Fprofile_engine.py` evidence. A fix goes back through step 8.\n\n   ```console\n   $ uv run scripts\u002Fbenchmark.py compare origin\u002Fmaster HEAD --format json --output \u003Cscratch>\u002Fbench.json --no-progress\n   ```\n\n## Report\n\nEnd with exactly these parts:\n\n- A table: agent | old → new version | what changed | searchable (record\n  count) | perf delta.\n- Each store left documented but unread, with the reason.\n- `CHANGES` entries the user-visible changes need, unless commit 8 landed\n  them.\n- Anything left undone, with the reason.\n","# Refreshing agent backends\n\nEach installed client version gets a schema-only manifest in\n`docs\u002F_observations\u002F\u003Cagent>\u002F\u003Cversion>.toml`, and the catalogue stamps in\n`src\u002Fagentgrep\u002Fstore_catalog\u002F` must match the newest one. A refresh records\nnew versions, catalogues what changed, ties each change to the version that\nfirst showed it, and proves search still works and is no slower.\n\n## Arguments\n\n`$ARGUMENTS` lists agent ids as `uv run scripts\u002Fobserve_stores.py --help`\nprints them; `--agent` takes one id, so run each command once per id. Empty\nmeans `--agent all`. \"In scope\" below means these agents.\n\n## Rules\n\n- Agent homes, `\u002Fmnt\u002Fc` included, are read-only: read files and run\n  `--version`, never another client subcommand. No `HOME=` overrides.\n- No prompt text, PII, or real home paths anywhere: manifests use `${HOME}`\n  tokens, docs pages keep their existing `~\u002F` form.\n- Never hand-edit a manifest, and never rewrite or delete one for a version\n  that is no longer installed. One that already leaks a path is the user's\n  call to scrub: report it.\n- Never push, tag, or open a PR. No backup branches.\n- If the user asked for `CHANGES`, its entries land last, in their own\n  `docs(CHANGES) \u003Cdescription>` commit under the existing unreleased\n  heading, with no version lead paragraph. Otherwise leave `CHANGES` alone\n  and list the entries it needs in the report.\n\n## Commits\n\nCommit in this order, skipping a kind with nothing in it. Subjects stay at\n50 characters or fewer, and bodies take the `why:`\u002F`what:` form from\n`.github\u002FWRITING.md`.\n\n1. `agentgrep(fix[observe]): ...` when the observer itself misbehaves,\n   with a test in `tests\u002Ftest_storage_observations.py`.\n2. `agentgrep(chore[catalog]): Refresh observed stamps`: every new\n   manifest, restamp, and version line. The body names each agent and\n   version recorded, and says earlier manifests stay untouched.\n3. `agentgrep(feat[catalog]): Catalogue \u003Cn> new stores`: new rows, the\n   `catalog_version` bump, and the touched agents re-observed at their\n   unchanged versions.\n4. `agentgrep(feat[\u003Cagent>]): ...` or `agentgrep(fix[\u003Cagent>]): ...`:\n   adapter changes with their fixture and test.\n5. `docs(backends): Correct claims the sweep disproved`.\n6. `docs(backends): Record storage changes by version`.\n7. `docs(backends): Observe \u003CAgent> \u003Cversion>` for a client that updates\n   mid-sweep: its manifest, restamp, version lines, and entry in one commit.\n8. `docs(CHANGES) \u003Cdescription>`, when the user asked for it.\n\n## Steps\n\n1. **Worktree.** Unless the user names a branch, branch from fresh\n   `origin\u002Fmaster` into a sibling worktree, with the session date as\n   `YYYYMMDD`. Work only there.\n\n   ```console\n   $ git fetch origin\n   ```\n\n   ```console\n   $ git worktree add --no-track -b storages-YYYYMMDD ..\u002Fagentgrep-storages-YYYYMMDD origin\u002Fmaster\n   ```\n\n2. **Baseline.** Exit 1 means drift. Table each in-scope agent's documented\n   version (newest manifest; `_\u003CAGENT>_OBSERVED_VERSION` in\n   `src\u002Fagentgrep\u002Fstore_catalog\u002F\u003Cagent>.py`, hyphens as underscores) against\n   the installed version this prints.\n\n   ```console\n   $ uv run scripts\u002Fobserve_stores.py check --agent \u003Cid>\n   ```\n\n   `windsurf` has no version probe, so `check` never reports its upgrades.\n   Read `windsurfVersion` from the newest\n   `~\u002F.windsurf-server\u002Fbin\u002F\u003Ccommit>\u002Fproduct.json`, and decide whether that\n   build wrote its catalogued data before stamping it. The first `check`\n   after the IDE probes landed reports each IDE once as drift from its\n   `unknown.toml`.\n\n3. **Observe** each in-scope agent whose version changed or that `check`\n   reports drifting. A new version writes a new manifest; drift at the same\n   version rewrites that version's file. A refusal to write `unknown.toml`\n   over a versioned manifest means the version probe failed: report it,\n   never work around it.\n\n   ```console\n   $ uv run scripts\u002Fobserve_stores.py observe --agent \u003Cid>\n   ```\n\n   A new version diffs against the previous manifest:\n\n   ```console\n   $ git diff --no-index docs\u002F_observations\u002F\u003Cid>\u002F\u003Cold>.toml docs\u002F_observations\u002F\u003Cid>\u002F\u003Cnew>.toml\n   ```\n\n   A same-version rewrite diffs against the commit:\n\n   ```console\n   $ git diff -- docs\u002F_observations\u002F\u003Cid>\u002F\u003Cversion>.toml\n   ```\n\n4. **Triage** every diff line:\n\n   - Before calling a key added or removed, check older manifests. The\n     observer samples each store's newest 40 files, so keys drift in and out\n     of the sample without a schema change.\n   - Version changed, shape didn't: restamp only.\n   - New unclaimed path: add a `catalog_only` row, or `private` when it\n     holds credentials or keys. A new home directory for the agent also goes\n     in its `AgentProbe.homes` in `scripts\u002Fobserve_stores.py`.\n   - Changed discriminator, keys, tables, or columns in a store agentgrep\n     reads: update the descriptor and adapter, add a redacted fixture under\n     `tests\u002Fsamples\u002F\u003Cagent>\u002F\u003Cstore_id>\u002F`, and show the new test failing\n     before it passes.\n   - `source_count \u003Cold> -> \u003Cnew>` on a store agentgrep reads means its\n     files moved, were renamed, or were recompressed, and search lost them.\n     Make discovery and the adapter follow them, or, when that needs a\n     design, list it first under what was left undone.\n   - New readable conversation store: give it the `primary_chat` or\n     `supplementary_chat` role, never `prompt_history`, so default search\n     does not open it.\n   - A file newer than its newest record is an empty draft session, not a\n     parse failure.\n   - Re-read the agent's backend page and each touched row's\n     `schema_notes`, and correct any claim the observation disproves.\n\n   Any descriptor add, removal, or reshape follows \"Adding or updating a\n   store\" in `docs\u002Fdev\u002Fstorage-catalog.md`, including its `upstream_ref`\n   (pinned to an upstream commit, `github.com\u002F\u003Corg>\u002F\u003Crepo>@\u003Csha>\u002F...`) or\n   `sample_record`. Bump `catalog_version` and re-observe the agent in that\n   same commit. If the unclaimed list now names the children of a directory,\n   the new row claims that directory by name without covering its children:\n   put a `\u003Cplaceholder>` or glob at that level of its pattern.\n\n5. **Restamp** only what you re-read:\n\n   - `_\u003CAGENT>_OBSERVED_VERSION`, keeping its prefix (`claude-code v...`).\n   - `_\u003CAGENT>_OBSERVED_AT`, `_CATALOG_CAPTURED_AT`, and the date in the\n     comment above the dated block, in\n     `src\u002Fagentgrep\u002Fstore_catalog\u002F_common.py`.\n   - The `observed_version` line wherever `docs\u002Fbackends\u002F\u003Cagent>.md` and\n     `docs\u002Fdev\u002Fstorage-catalog.md` carry one. Windsurf's page is\n     `docs\u002Fbackends\u002Funsupported\u002Fwindsurf.md`.\n\n   An agent that is not installed, or whose version you could not read,\n   keeps its version and date: move its `_OBSERVED_AT` out of the dated\n   block with a comment saying why, as Windsurf's is. When the gate fails on\n   a row that keeps its own `observed_at`, re-verify that row on disk before\n   pointing it at the module constant.\n\n6. **Version the change** under `## Changes by version` on the agent's page,\n   adding that section, with the opening paragraph other pages use, where it\n   is missing. Each `### \u003Cnew>` entry opens \"Seen in \u003Cnew> (YYYY-MM-DD);\n   absent in \u003Cold> (YYYY-MM-DD).\" Versions between two observations were\n   never seen, so a change landed after `\u003Cold>` and by `\u003Cnew>`; never name an\n   exact version. With no readable version, the entry is `### Observed\n   YYYY-MM-DD`, bracketed \"seen \u003Cdate>, absent \u003Cdate>\", naming any build\n   version you could read. A version with no storage change still gets an\n   entry saying so. A store agentgrep does not read is \"not searched\", never\n   \"excluded\".\n\n7. **Prove search** for every touched agent id except `windsurf`, which has\n   no search surface; check it through its docs page. Pick `\u003Cterm>` from a\n   store you changed, and keep it out of commits and the report.\n\n   ```console\n   $ uv run agentgrep find --agent \u003Cid> --json\n   ```\n\n   ```console\n   $ uv run agentgrep search --agent \u003Cid> --scope all --exhaustive --json \u003Cterm>\n   ```\n\n   Pass means records returned, `summary.coverage.sources_failed` is 0, and\n   `summary.diagnostics` has nothing new. For a field you started reading\n   (model, cwd, title), returned records must carry real values; a key in\n   the output proves nothing. Then run the Layer 0 MCP smoke from the\n   `testing-mcp-with-cli-agents` skill: a throwaway FastMCP stdio client in\n   the scratchpad doing `tools\u002Flist` plus one `search`.\n\n8. **Gate, then commit.** Before every commit: `rm -rf docs\u002F_build`, the\n   gates in `.github\u002FCONTRIBUTING.md`, `just test-all`, and\n   `just build-docs`. When `FORCE_COLOR` is set, run tests under\n   `NO_COLOR=1`. Read each command's own pass\u002Ffail line, never a `| tail` or\n   `| grep` of it. `just build-docs` lines starting `storage observations:`\n   mean a stamp and its newest manifest disagree. Then stage and scan the\n   added lines; it must print nothing. It catches encoded project\n   directories such as `home-\u003Cuser>-work-...` and `%2Fhome%2F...`.\n\n   ```console\n   $ git diff --cached | grep -nE \"^\\+.*(\u002Fhome\u002F|\u002Fmnt\u002F|\u002FUsers\u002F|home-$USER-|%2F(home|Users)%2F)\"\n   ```\n\n   Before the last commit, `check` exits 0 for every in-scope agent; a\n   client that updated mid-sweep gets commit 7.\n\n9. **Measure** after the last commit, with every configured bench and its\n   configured runs and warmup, which each bench's timeout is sized for; the\n   `benchmark` skill covers flags and `analyze`. Default `search` must not\n   slow beyond run-to-run noise; explain any remaining delta with\n   `scripts\u002Fprofile_engine.py` evidence. A fix goes back through step 8.\n\n   ```console\n   $ uv run scripts\u002Fbenchmark.py compare origin\u002Fmaster HEAD --format json --output \u003Cscratch>\u002Fbench.json --no-progress\n   ```\n\n## Report\n\nEnd with exactly these parts:\n\n- A table: agent | old → new version | what changed | searchable (record\n  count) | perf delta.\n- Each store left documented but unread, with the reason.\n- `CHANGES` entries the user-visible changes need, unless commit 8 landed\n  them.\n- Anything left undone, with the reason.\n",{"name":5,"description":7,"argument-hint":24},"[agent ...]","master","b4cc32321965fa6c908b7d2d7e0650bbee9f85c4",".agents\u002Fskills\u002Frefresh-backends",1,"https:\u002F\u002Fgithub.com\u002Ftony\u002Fagentgrep","https:\u002F\u002Fagentgrep.org","https:\u002F\u002Fgithub.com\u002Ftony","https:\u002F\u002Ftony.sh","lighthouse keeper","2026-11-15 04:00:32",{"source":8,"skillName":5,"commands":36,"prompt":46,"zipAvailable":47},[37,40,43],{"tool":38,"command":39},"npx","npx skills add tony\u002Fagentgrep --skill refresh-backends",{"tool":41,"command":42},"bunx","bunx skills add tony\u002Fagentgrep --skill refresh-backends",{"tool":44,"command":45},"pnpm","pnpm dlx skills add tony\u002Fagentgrep --skill refresh-backends","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Ftony\u002Fagentgrep\u002Fblob\u002Fmaster\u002F.agents\u002Fskills\u002Frefresh-backends\u002FSKILL.md",true,[],[50,56,62,69],{"skillNo":51,"name":52,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":53,"categories":54,"keywords":55,"publishedAt":19},"01M72E6AC6YTG31RKX4EM5EE0D","profile","https:\u002F\u002Fgithub.com\u002Ftony\u002Fagentgrep\u002Fblob\u002Fmaster\u002F.claude\u002Fskills\u002Fprofile\u002FSKILL.md",[],[],{"skillNo":57,"name":58,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":59,"categories":60,"keywords":61,"publishedAt":19},"01M72E695MZ9S0VB6DJH5EDRC2","benchmark","https:\u002F\u002Fgithub.com\u002Ftony\u002Fagentgrep\u002Fblob\u002Fmaster\u002F.claude\u002Fskills\u002Fbenchmark\u002FSKILL.md",[],[],{"skillNo":63,"name":64,"nameZh":6,"description":65,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":66,"categories":67,"keywords":68,"publishedAt":19},"01M72E657SDH6GA08WGP5ES057","textual-non-blocking-pump","Use when writing, reviewing, or auditing agentgrep TUI code (src\u002Fagentgrep\u002Fui\u002F) for anything that could freeze the Textual message pump — keystrokes\u002Fspinner\u002Fcancel hang, a handler that blocks, a slow filter\u002Ffind\u002Fdetail render, \"the UI froze\", a new pump method\u002Ftimer\u002Fworker, or revising the ADR 0011 non-blocking safeguards.","https:\u002F\u002Fgithub.com\u002Ftony\u002Fagentgrep\u002Fblob\u002Fmaster\u002F.agents\u002Fskills\u002Ftextual-non-blocking-pump\u002FSKILL.md",[],[],{"skillNo":70,"name":71,"nameZh":6,"description":72,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":73,"categories":74,"keywords":75,"publishedAt":19},"01M72E5ZBCB33ZS1SBS0E2WBFK","testing-mcp-with-cli-agents","Test an MCP server by driving real CLI agents (Claude, Codex, Cursor, Gemini, Grok, agy) against it — isolating each CLI's config and the server's own backend state instead of trusting unit tests alone. Use this whenever verifying MCP-server behavior end-to-end, checking that a local branch or checkout works across installed agent CLIs, comparing trunk-vs-branch behavior, driving an interactive agent TUI to exercise approval flows or cancellation, or reproducing a bug through a live client. Reach for it even when the user only says \"test the MCP\", \"does the branch work in the agents\", \"drive the CLI to call the tool\", or \"check it across Codex\u002FGemini\u002FCursor\".","https:\u002F\u002Fgithub.com\u002Ftony\u002Fagentgrep\u002Fblob\u002Fmaster\u002F.agents\u002Fskills\u002Ftesting-mcp-with-cli-agents\u002FSKILL.md",[],[]]