[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M7CW5P474EM4RSFSMQPJ7XW1":3,"skill-same:01M7CW5P474EM4RSFSMQPJ7XW1":71},{"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":23,"commitSha":24,"skillDirectory":25,"forksCount":13,"repositoryHtmlUrl":26,"repositoryHomepageUrl":6,"ownerHtmlUrl":27,"ownerBlogUrl":6,"ownerDescription":6,"updatedAt":28,"install":29,"resources":42},"01M7CW5P474EM4RSFSMQPJ7XW1","security-permissions",null,"Risk-graded permission audit - wraps native \u002Ffewer-permission-prompts with allow\u002Fask\u002Fdeny tiers","cooneycw\u002Fskillc","cooneycw","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F36539213?v=4","01M7CVQ6GTEEQ605CAJ92QMJ15","01M7CVQ6HGVRC336M6PCCB4FYE",0,"MIT","https:\u002F\u002Fgithub.com\u002Fcooneycw\u002Fskillc\u002Fblob\u002Fmain\u002Ftests\u002Ffixtures\u002Fprofile-cpp-codex-flow-auto-87ecb097\u002Fcodex\u002Fskills\u002Fsecurity-permissions\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: \"security-permissions\"\ndescription: \"Risk-graded permission audit - wraps native \u002Ffewer-permission-prompts with allow\u002Fask\u002Fdeny tiers\"\n---\n\u003C!-- GENERATED by claude-power-pack - scripts\u002Fcodex-skill-sync.py; edit .claude\u002Fcommands\u002Fsecurity\u002Fpermissions.md instead -->\n\n## Codex harness adaptations\n\nGenerated from a Claude Code command. Where the procedure references these Claude-only surfaces, adapt as follows:\n\n- MCP tools: use the MCP servers configured in `~\u002F.codex\u002Fconfig.toml`, or fall back to the referenced repo scripts and CLI entry points.\n- Helper scripts are bundled under `scripts\u002F` in this skill directory (byte-identical copies from the claude-power-pack checkout). Commands run them as `\"${CPP_SKILL_DIR:-\u002Fdev\u002Fnull}\"\u002Fscripts\u002F\u003Cname>`: `CPP_SKILL_DIR` is the absolute directory containing THIS SKILL.md - the location the skills catalog gives, with any alias such as `r0` expanded. A shell resolves a bare `scripts\u002F\u003Cname>` against your working directory (the target project), never this skill, so substitute that directory, or set it in the same command (`CPP_SKILL_DIR=\u002Fabs\u002Fskill\u002Fdir; ...`) - each command is a fresh shell. \"Not a directory\" on `\u002Fdev\u002Fnull\u002Fscripts\u002F\u003Cname>` (exit 126, or 127 in some shells) means `CPP_SKILL_DIR` is unset: set it and re-run; never run the target project's own `scripts\u002F\u003Cname>` in its place. Read every \"exit 127\" fallback in this skill as \"exit 127, or 126 Not a directory\": set the variable and re-run before falling back. A helper whose verdict gates a step and that prints no verdict line after the whole fallback chain did not run - for the finish gate, no `FLOW_FINISH_GATE:` verdict line: STOP, never read it as skipped, never commit on it (issue #1547). A bare `scripts\u002F\u003Cname>` inside a command is a CPP-checkout operation that the text says so about; in prose, `scripts\u002F\u003Cname>` NAMES the bundled helper. Some helpers expect sibling repo resources, so prefer a full checkout when one is available.\n\n# Security: Permissions Audit\n\nA risk-graded wrapper around Claude Code's native **`\u002Ffewer-permission-prompts`**.\nThe native skill does a binary read-only \u002F skip split and emits an allowlist only.\nThis command adds `scripts\u002Fclassify-tool-risk.py` to sort every observed tool call\ninto risk tiers, then proposes a full **allow \u002F ask \u002F deny** policy.\n\n> Use the native `\u002Ffewer-permission-prompts` for the quick read-only allowlist.\n> Use this when you also want the mutation and danger tiers surfaced and guarded\n> (it never edits `permissions.ask` \u002F `permissions.deny` unless you asked for the\n> graded policy, which this command does by design).\n\n## What it produces\n\n- **allow** - read-only commands not already auto-allowed (e.g. `git fetch`) +\n  the *safe, reversible* subset of local writes (`git add`, `git commit`,\n  `mkdir`, `git worktree add`) + read-only MCP tools seen >= 3 times.\n- **ask** - the recoverable-but-dangerous tier, gated *even under auto-accept*:\n  `git push --force`, `git reset --hard`, `rm -rf`.\n- **deny** - secrets that must never be auto-edited: `.env`, `*.pem`, `*.key`,\n  `id_rsa*`.\n\n## Steps\n\n1. **Run the graded classifier** over recent transcripts:\n   ```bash\n   python3 \"${CPP_SKILL_DIR:-\u002Fdev\u002Fnull}\"\u002Fscripts\u002Fclassify-tool-risk.py --limit 50\n   ```\n   (Run from the CPP checkout. It scans `~\u002F.claude\u002Fprojects\u002F**\u002F*.jsonl` across all\n   projects, not just the current one.)\n\n2. **Present the tiers** as a table (tier, count, examples, verdict). Note that\n   the READONLY-AUTO tier is already covered by Claude Code and needs no rule -\n   do not propose those (see the native skill's auto-allow list).\n\n3. **Propose the policy from the tiers:**\n   - `allow`: READONLY-ADDABLE + the SAFE subset of WRITE-LOCAL. Exclude anything\n     with a data-loss mode - `git checkout`, `git rm`, `git branch -D`,\n     `git worktree remove`, `git reset` classify as DESTRUCTIVE, not allow.\n     CODE-EXEC \u002F DUAL-USE-NET \u002F SCRIPT-EXEC are **never** allowlisted (arbitrary\n     execution or ambiguous network I\u002FO).\n   - `ask`: the DESTRUCTIVE tier (force push, `reset --hard`, `rm -rf`) so it\n     prompts even under auto-accept.\n   - `deny`: secret files (`.env`, `**\u002F.env`, `*.pem`, `*.key`, `id_rsa*`).\n\n4. **Merge into the project `.claude\u002Fsettings.json`** (NOT `settings.local.json`,\n   NOT `~\u002F.claude\u002Fsettings.json`): preserve existing keys and entries,\n   de-duplicate, never remove or reorder. Precedence is **deny > ask > allow**, so\n   a secret `deny` overrides a broad `Edit(**)` allow.\n\n5. **Report**: what was added to each tier, what was already covered by Claude\n   Code's auto-allow (so no rule was needed), and what was excluded and why\n   (e.g. dropped `python3`\u002F`uv`\u002F`make` - arbitrary code execution; dropped\n   `curl`\u002F`gh api` - dual-use network).\n\n## File-edit scope (the git-tracked nuance)\n\nTo auto-allow file edits \"only within the git commit list\": Claude Code\npermissions are path-glob based, not git-aware, so this cannot be expressed\ndirectly. Two implementations:\n\n- **Approximation (default):** `Edit(**)` \u002F `Write(**)` for the repo tree plus a\n  `deny` list for sensitive gitignored files (`.env`, keys).\n- **Exact (offer):** a `PreToolUse` hook that runs `git check-ignore` on the edit\n  target and blocks ignored paths. This literally scopes edits to tracked files.\n  Offered, not applied by default.\n\n## Notes\n\n- Wraps, does not replace, the native `\u002Ffewer-permission-prompts` (a built-in\n  Claude Code skill, not a CPP skill - which is why its graded extension lives\n  in CPP).\n- Classifier: `scripts\u002Fclassify-tool-risk.py` (the **canonical** risk taxonomy).\n  Re-run anytime for a fresh view.\n- Complements the **real-time** side: the `#482` permission-census hook\n  (`scripts\u002Fhook-permission-census.sh`) captures prompts as they happen and\n  vendors the same taxonomy inline. This command is the **retroactive** side\n  (mine all transcripts, propose a policy in one pass). `scripts\u002Ftool-risk-drift.py`\n  guards the two taxonomies' safety-critical sets in sync.\n- The destructive `ask`\u002F`deny` guards are more protective in\n  `~\u002F.claude\u002Fsettings.json` (global, every project); the allowlist is\n  project-scoped by design. Offer to promote the guards to global.\n","\u003C!-- GENERATED by claude-power-pack - scripts\u002Fcodex-skill-sync.py; edit .claude\u002Fcommands\u002Fsecurity\u002Fpermissions.md instead -->\n\n## Codex harness adaptations\n\nGenerated from a Claude Code command. Where the procedure references these Claude-only surfaces, adapt as follows:\n\n- MCP tools: use the MCP servers configured in `~\u002F.codex\u002Fconfig.toml`, or fall back to the referenced repo scripts and CLI entry points.\n- Helper scripts are bundled under `scripts\u002F` in this skill directory (byte-identical copies from the claude-power-pack checkout). Commands run them as `\"${CPP_SKILL_DIR:-\u002Fdev\u002Fnull}\"\u002Fscripts\u002F\u003Cname>`: `CPP_SKILL_DIR` is the absolute directory containing THIS SKILL.md - the location the skills catalog gives, with any alias such as `r0` expanded. A shell resolves a bare `scripts\u002F\u003Cname>` against your working directory (the target project), never this skill, so substitute that directory, or set it in the same command (`CPP_SKILL_DIR=\u002Fabs\u002Fskill\u002Fdir; ...`) - each command is a fresh shell. \"Not a directory\" on `\u002Fdev\u002Fnull\u002Fscripts\u002F\u003Cname>` (exit 126, or 127 in some shells) means `CPP_SKILL_DIR` is unset: set it and re-run; never run the target project's own `scripts\u002F\u003Cname>` in its place. Read every \"exit 127\" fallback in this skill as \"exit 127, or 126 Not a directory\": set the variable and re-run before falling back. A helper whose verdict gates a step and that prints no verdict line after the whole fallback chain did not run - for the finish gate, no `FLOW_FINISH_GATE:` verdict line: STOP, never read it as skipped, never commit on it (issue #1547). A bare `scripts\u002F\u003Cname>` inside a command is a CPP-checkout operation that the text says so about; in prose, `scripts\u002F\u003Cname>` NAMES the bundled helper. Some helpers expect sibling repo resources, so prefer a full checkout when one is available.\n\n# Security: Permissions Audit\n\nA risk-graded wrapper around Claude Code's native **`\u002Ffewer-permission-prompts`**.\nThe native skill does a binary read-only \u002F skip split and emits an allowlist only.\nThis command adds `scripts\u002Fclassify-tool-risk.py` to sort every observed tool call\ninto risk tiers, then proposes a full **allow \u002F ask \u002F deny** policy.\n\n> Use the native `\u002Ffewer-permission-prompts` for the quick read-only allowlist.\n> Use this when you also want the mutation and danger tiers surfaced and guarded\n> (it never edits `permissions.ask` \u002F `permissions.deny` unless you asked for the\n> graded policy, which this command does by design).\n\n## What it produces\n\n- **allow** - read-only commands not already auto-allowed (e.g. `git fetch`) +\n  the *safe, reversible* subset of local writes (`git add`, `git commit`,\n  `mkdir`, `git worktree add`) + read-only MCP tools seen >= 3 times.\n- **ask** - the recoverable-but-dangerous tier, gated *even under auto-accept*:\n  `git push --force`, `git reset --hard`, `rm -rf`.\n- **deny** - secrets that must never be auto-edited: `.env`, `*.pem`, `*.key`,\n  `id_rsa*`.\n\n## Steps\n\n1. **Run the graded classifier** over recent transcripts:\n   ```bash\n   python3 \"${CPP_SKILL_DIR:-\u002Fdev\u002Fnull}\"\u002Fscripts\u002Fclassify-tool-risk.py --limit 50\n   ```\n   (Run from the CPP checkout. It scans `~\u002F.claude\u002Fprojects\u002F**\u002F*.jsonl` across all\n   projects, not just the current one.)\n\n2. **Present the tiers** as a table (tier, count, examples, verdict). Note that\n   the READONLY-AUTO tier is already covered by Claude Code and needs no rule -\n   do not propose those (see the native skill's auto-allow list).\n\n3. **Propose the policy from the tiers:**\n   - `allow`: READONLY-ADDABLE + the SAFE subset of WRITE-LOCAL. Exclude anything\n     with a data-loss mode - `git checkout`, `git rm`, `git branch -D`,\n     `git worktree remove`, `git reset` classify as DESTRUCTIVE, not allow.\n     CODE-EXEC \u002F DUAL-USE-NET \u002F SCRIPT-EXEC are **never** allowlisted (arbitrary\n     execution or ambiguous network I\u002FO).\n   - `ask`: the DESTRUCTIVE tier (force push, `reset --hard`, `rm -rf`) so it\n     prompts even under auto-accept.\n   - `deny`: secret files (`.env`, `**\u002F.env`, `*.pem`, `*.key`, `id_rsa*`).\n\n4. **Merge into the project `.claude\u002Fsettings.json`** (NOT `settings.local.json`,\n   NOT `~\u002F.claude\u002Fsettings.json`): preserve existing keys and entries,\n   de-duplicate, never remove or reorder. Precedence is **deny > ask > allow**, so\n   a secret `deny` overrides a broad `Edit(**)` allow.\n\n5. **Report**: what was added to each tier, what was already covered by Claude\n   Code's auto-allow (so no rule was needed), and what was excluded and why\n   (e.g. dropped `python3`\u002F`uv`\u002F`make` - arbitrary code execution; dropped\n   `curl`\u002F`gh api` - dual-use network).\n\n## File-edit scope (the git-tracked nuance)\n\nTo auto-allow file edits \"only within the git commit list\": Claude Code\npermissions are path-glob based, not git-aware, so this cannot be expressed\ndirectly. Two implementations:\n\n- **Approximation (default):** `Edit(**)` \u002F `Write(**)` for the repo tree plus a\n  `deny` list for sensitive gitignored files (`.env`, keys).\n- **Exact (offer):** a `PreToolUse` hook that runs `git check-ignore` on the edit\n  target and blocks ignored paths. This literally scopes edits to tracked files.\n  Offered, not applied by default.\n\n## Notes\n\n- Wraps, does not replace, the native `\u002Ffewer-permission-prompts` (a built-in\n  Claude Code skill, not a CPP skill - which is why its graded extension lives\n  in CPP).\n- Classifier: `scripts\u002Fclassify-tool-risk.py` (the **canonical** risk taxonomy).\n  Re-run anytime for a fresh view.\n- Complements the **real-time** side: the `#482` permission-census hook\n  (`scripts\u002Fhook-permission-census.sh`) captures prompts as they happen and\n  vendors the same taxonomy inline. This command is the **retroactive** side\n  (mine all transcripts, propose a policy in one pass). `scripts\u002Ftool-risk-drift.py`\n  guards the two taxonomies' safety-critical sets in sync.\n- The destructive `ask`\u002F`deny` guards are more protective in\n  `~\u002F.claude\u002Fsettings.json` (global, every project); the allowlist is\n  project-scoped by design. Offer to promote the guards to global.\n",{"name":5,"description":7},"main","26e686aea67467e80e66f7d622950d0702c0a092","tests\u002Ffixtures\u002Fprofile-cpp-codex-flow-auto-87ecb097\u002Fcodex\u002Fskills\u002Fsecurity-permissions","https:\u002F\u002Fgithub.com\u002Fcooneycw\u002Fskillc","https:\u002F\u002Fgithub.com\u002Fcooneycw","2026-11-15 05:18:27",{"source":8,"skillName":5,"commands":30,"prompt":40,"zipAvailable":41},[31,34,37],{"tool":32,"command":33},"npx","npx skills add cooneycw\u002Fskillc --skill security-permissions",{"tool":35,"command":36},"bunx","bunx skills add cooneycw\u002Fskillc --skill security-permissions",{"tool":38,"command":39},"pnpm","pnpm dlx skills add cooneycw\u002Fskillc --skill security-permissions","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fcooneycw\u002Fskillc\u002Fblob\u002Fmain\u002Ftests\u002Ffixtures\u002Fprofile-cpp-codex-flow-auto-87ecb097\u002Fcodex\u002Fskills\u002Fsecurity-permissions\u002FSKILL.md",true,[43,49,55,59,63,67],{"path":44,"resourceType":45,"resourceTypeDesc":46,"sizeBytes":47,"sourceHtmlUrl":48},"PROVENANCE.md","REFERENCE","参考文档",579,"https:\u002F\u002Fgithub.com\u002Fcooneycw\u002Fskillc\u002Fblob\u002Fmain\u002Ftests\u002Ffixtures\u002Fprofile-cpp-codex-flow-auto-87ecb097\u002Fcodex\u002Fskills\u002Fsecurity-permissions\u002FPROVENANCE.md",{"path":50,"resourceType":51,"resourceTypeDesc":52,"sizeBytes":53,"sourceHtmlUrl":54},"scripts\u002Fclassify-tool-risk.py","SCRIPT","脚本",9863,"https:\u002F\u002Fgithub.com\u002Fcooneycw\u002Fskillc\u002Fblob\u002Fmain\u002Ftests\u002Ffixtures\u002Fprofile-cpp-codex-flow-auto-87ecb097\u002Fcodex\u002Fskills\u002Fsecurity-permissions\u002Fscripts\u002Fclassify-tool-risk.py",{"path":56,"resourceType":51,"resourceTypeDesc":52,"sizeBytes":57,"sourceHtmlUrl":58},"scripts\u002Ffriction-log.sh",7271,"https:\u002F\u002Fgithub.com\u002Fcooneycw\u002Fskillc\u002Fblob\u002Fmain\u002Ftests\u002Ffixtures\u002Fprofile-cpp-codex-flow-auto-87ecb097\u002Fcodex\u002Fskills\u002Fsecurity-permissions\u002Fscripts\u002Ffriction-log.sh",{"path":60,"resourceType":51,"resourceTypeDesc":52,"sizeBytes":61,"sourceHtmlUrl":62},"scripts\u002Fhook-permission-census.sh",21117,"https:\u002F\u002Fgithub.com\u002Fcooneycw\u002Fskillc\u002Fblob\u002Fmain\u002Ftests\u002Ffixtures\u002Fprofile-cpp-codex-flow-auto-87ecb097\u002Fcodex\u002Fskills\u002Fsecurity-permissions\u002Fscripts\u002Fhook-permission-census.sh",{"path":64,"resourceType":51,"resourceTypeDesc":52,"sizeBytes":65,"sourceHtmlUrl":66},"scripts\u002FSHA256SUMS",924,"https:\u002F\u002Fgithub.com\u002Fcooneycw\u002Fskillc\u002Fblob\u002Fmain\u002Ftests\u002Ffixtures\u002Fprofile-cpp-codex-flow-auto-87ecb097\u002Fcodex\u002Fskills\u002Fsecurity-permissions\u002Fscripts\u002FSHA256SUMS",{"path":68,"resourceType":51,"resourceTypeDesc":52,"sizeBytes":69,"sourceHtmlUrl":70},"scripts\u002Ftool-risk-drift.py",3108,"https:\u002F\u002Fgithub.com\u002Fcooneycw\u002Fskillc\u002Fblob\u002Fmain\u002Ftests\u002Ffixtures\u002Fprofile-cpp-codex-flow-auto-87ecb097\u002Fcodex\u002Fskills\u002Fsecurity-permissions\u002Fscripts\u002Ftool-risk-drift.py",[72,79,85,92,97,102],{"skillNo":73,"name":74,"nameZh":6,"description":75,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":76,"categories":77,"keywords":78,"publishedAt":18},"01M7CW6RKZX2298AZDERY3EZQR","tiny-check","Run the tiny project check","https:\u002F\u002Fgithub.com\u002Fcooneycw\u002Fskillc\u002Fblob\u002Fmain\u002Ftests\u002Ffixtures\u002Fprofile-install\u002Fskills\u002Ftiny-check\u002FSKILL.md",[],[],{"skillNo":80,"name":74,"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},"01M7CW6Q781BQX42JV5GE8YSCV","A synthetic gate selection fixture.","https:\u002F\u002Fgithub.com\u002Fcooneycw\u002Fskillc\u002Fblob\u002Fmain\u002Ftests\u002Ffixtures\u002Fprofile-install-synthetic\u002Fskills\u002Ftiny-check\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},"01M7CW6MZM805TYJFVYYFSV7AB","flow-check","Use when you are asked to run a project's quality checks (tests, lint, type checks, security scan): runs them in one step and reports what failed, without committing.","https:\u002F\u002Fgithub.com\u002Fcooneycw\u002Fskillc\u002Fblob\u002Fmain\u002Ftests\u002Ffixtures\u002Fprofile-cpp-codex-flow-check-resume-v1\u002Fcodex\u002Fskills\u002Fflow-check\u002FSKILL.md",[],[],{"skillNo":93,"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":94,"categories":95,"keywords":96,"publishedAt":18},"01M7CW6J87T788K88H0M5G21DZ","https:\u002F\u002Fgithub.com\u002Fcooneycw\u002Fskillc\u002Fblob\u002Fmain\u002Ftests\u002Ffixtures\u002Fprofile-cpp-codex-flow-check-ea6dbfa\u002Fcodex\u002Fskills\u002Fflow-check\u002FSKILL.md",[],[],{"skillNo":98,"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":99,"categories":100,"keywords":101,"publishedAt":18},"01M7CW6FE77JB8Z0ZFNH2GJVNE","https:\u002F\u002Fgithub.com\u002Fcooneycw\u002Fskillc\u002Fblob\u002Fmain\u002Ftests\u002Ffixtures\u002Fprofile-cpp-codex-flow-check-86641e8\u002Fcodex\u002Fskills\u002Fflow-check\u002FSKILL.md",[],[],{"skillNo":103,"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":104,"categories":105,"keywords":106,"publishedAt":18},"01M7CW6CD3KN3S6D3QRNYN50JZ","https:\u002F\u002Fgithub.com\u002Fcooneycw\u002Fskillc\u002Fblob\u002Fmain\u002Ftests\u002Ffixtures\u002Fprofile-cpp-codex-flow-check-652c0118\u002Fcodex\u002Fskills\u002Fflow-check\u002FSKILL.md",[],[]]