[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M7C97EE1Y5W95T3ZJEXFS57G":3,"skill-same:01M7C97EE1Y5W95T3ZJEXFS57G":54},{"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":29,"repositoryHtmlUrl":30,"repositoryHomepageUrl":6,"ownerHtmlUrl":31,"ownerBlogUrl":6,"ownerDescription":32,"updatedAt":33,"install":34,"resources":47},"01M7C97EE1Y5W95T3ZJEXFS57G","safe-execution-edho-ferdian",null,"Mechanical gates for agent runs: investigate before the first edit, guard destructive commands, freeze write scope for autonomous or parallel agents, block \"done\" until memory is updated. Use for autonomous or production work, or \"jangan sampai kehapus\", \"agent-nya nulis di luar scope\", \"pastiin dia ngecek dulu\".","edhoferdian\u002FEEF","edhoferdian","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F64404158?v=4","01M7C8538ZVVZWZEN1AW2MQXCV","Edho F","01M7C8539NS56BMJ4GXN8A47TQ",22,"MIT","https:\u002F\u002Fgithub.com\u002Fedhoferdian\u002FEEF\u002Fblob\u002Fmain\u002F.zcode\u002Fskills\u002Fsafe-execution-edho-ferdian\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: safe-execution-edho-ferdian\ndescription: >-\n  Mechanical gates for agent runs: investigate before the first edit, guard\n  destructive commands, freeze write scope for autonomous or parallel\n  agents, block \"done\" until memory is updated. Use for autonomous or\n  production work, or \"jangan sampai kehapus\", \"agent-nya nulis di luar\n  scope\", \"pastiin dia ngecek dulu\".\nmetadata:\n  last_reviewed: \"2026-10-27\"\n---\n\n# Safe Execution — Edho Ferdian Mode\n\n## Why this exists: reasoning gates cannot catch this class\n\nThis ecosystem is already dense with **reasoning gates** — the 8-gate\nReflection block, the Critique-Correction Loop, EDHO SCAN, the Output\nScorecard. Every one of them asks the model about its own work, after the\nwork is done.\n\nThat has a known ceiling. Asking a model \"did you violate anything?\"\nreliably returns \"no\" — self-evaluation shares the biases that produced the\noutput. What changes behaviour is being made to **produce a fact**: asked\nto list every file that imports a module, the model has to actually run\nGrep and Read, and the resulting context changes what it writes next.\nInternal A\u002FB measurements on gated vs ungated agents put the gap at\nroughly +2.25 points on a 10-point rubric across two tasks — a small\nsample, quoted as directional evidence rather than proof.\n\nSo the gates here are deliberately **not** reasoning gates. Each one either\ndemands an artifact or checks a machine-verifiable fact.\n\n## Gate 1 — Fact-forcing, before the first touch of a file\n\nThree stages, in order. Stopping at DENY is what most guards do and it is\nthe least useful part:\n\n```\nDENY  — block the first Edit\u002FWrite\u002FBash attempt against a target\nFORCE — state exactly which facts must be produced\nALLOW — permit the retry once those facts are on the table\n```\n\n**Before editing an existing file**, produce:\n1. Every file that imports or requires it (search the tree — Grep\u002FGlob; if\n   `salak` is on PATH, `salak query importers \u003Cfile>` and, for Go,\n   `salak query callers \u003Cnode>` give the edges with provenance — exit 2\n   means not tracked, so Grep instead; see\n   `dev-kickoff-edho-ferdian\u002Freferences\u002Fsalak-integration.md`).\n2. The public functions or classes this change affects.\n3. If it reads or writes data files: the field names, structure, and date\n   format — using redacted or synthetic values, never raw production data.\n4. The user's current instruction, quoted verbatim.\n\n**Before creating a new file**, produce:\n1. The file(s) and line(s) that will call it.\n2. Confirmation that no existing file already serves that purpose.\n3. Data-shape facts, as above, if it touches data files.\n4. The user's current instruction, quoted verbatim.\n\nFact 3 is the one people skip and the one that pays. In both A\u002FB trials\nthe ungated agent assumed ISO-8601 dates while the real data used\n`%Y\u002F%m\u002F%d %H:%M`. Checking the actual shape kills that entire bug class.\n\nFact 4 looks like ceremony and is not: re-reading the literal instruction\nbefore acting is the cheapest defence against the drift where an agent, six\ntool calls deep, is solving an adjacent problem it invented.\n\n**Scale the facts to the change, not to the file.** Read the edit itself\n(old and new text) before choosing which facts it owes:\n\n| The edit… | Owes |\n|---|---|\n| changes only comments, whitespace or formatting — same code once those are stripped | nothing; proceed |\n| changes a body, but no declaration that other files can see | the call sites that rely on this behaviour (narrower than fact 1), plus fact 4 |\n| touches the public surface — an exported\u002F`pub`\u002Fcapitalised name, a signature, a module's export list, a whole-file Write | fact 1 and fact 2 in full |\n| touches data — a format, schema, query, date\u002Ftime handling, file or network I\u002FO | fact 3, whatever else it owes |\n| cannot be classified with confidence | every fact; uncertainty never earns a discount |\n\nOver-matching is fine (a body inside an exported object counts as surface);\nunder-matching is the failure. This keeps the gate expensive exactly where\nthe A\u002FB bugs lived — data shape and public contracts — and nearly free for\nthe edits that never carried them.\n\nScope: **first touch per target, not every edit.** Once the facts are\npresented, further edits to that same file proceed freely. A gate that\nfires on every call gets disabled by lunchtime.\n\n## Gate 2 — Destructive commands, every single time\n\nUnlike Gate 1, this one never gets a free pass after the first hit.\n\nWatched: `rm -rf` (especially against `\u002F`, `~`, or a project root),\n`git reset --hard`, `git push --force`, `git checkout .`, `git clean -fd`,\n`DROP TABLE` \u002F `DROP DATABASE`, `docker system prune`, `kubectl delete`,\n`chmod 777`, `sudo rm`, `npm publish`, and anything carrying `--no-verify`.\n\nOn a hit, produce before running:\n1. Every file or dataset the command will modify or delete.\n2. A one-line rollback procedure.\n3. The user's current instruction, quoted verbatim.\n\nIf step 2 cannot be written, that is the answer: do not run it.\n\nPowerShell has the same list under other names: `Remove-Item -Recurse`\n(and `rm`\u002F`ri`\u002F`del`\u002F`rd` with `-Recurse` or `\u002Fs`), `Format-Volume`,\n`Clear-Disk`, `Clear-RecycleBin`. In Claude Code, the hook in the appendix\nturns every hit into a permission prompt the user must answer — even in\nauto mode — so the three facts above are what they read before approving.\n\n`--no-verify` is on the list deliberately — it is the mechanical form of\n\"skip the checks\", and it is exactly what an agent under time pressure\nreaches for.\n\n**Runaway searches are destructive too** — of time and of the machine. A\nrecursive search starting at a filesystem, drive, or home root (`find \u002F`,\n`find ~`, `rg x \u002F`, `Get-ChildItem -Recurse C:\\`, `dir \u002Fs C:\\`) never \u003C!-- fs-search-ok: names the banned forms -->\nneeds the three-step justification above: it is simply refused. On Windows\nit runs for hours and outlives the tool call as an orphaned process — this\necosystem's own agent wrappers once spawned six `find.exe` scans hunting for\na `SKILL.md`, some alive for ten hours. Search a known directory instead;\nif the target has no known location, ask. The Claude Code binding is\n`hooks\u002Fblock-fs-wide-search.js` in this skill (see the appendix).\n\nNever build a shell command by interpolating a string. Pass the executable\nand its arguments as separate entries with `shell: false` — string\ninterpolation is how a filename with a space or a quote becomes an\narbitrary command.\n\n## Gate 3 — Freeze: scope the write surface\n\nFreeze Mode locks Write\u002FEdit to a named\nsubtree; reads stay unrestricted. Anything outside is refused with an\nexplanation rather than silently allowed.\n\nThis is the directly load-bearing gate for how this ecosystem actually\nworks. Its own porting effort dispatches several Sonnet agents in parallel,\neach owning a slice of the tree — and the kelompok 3 batch found real\ndamage of exactly this shape afterwards (an agent leaving stale pointers\ninto out-of-scope paths across files outside its brief, fixed by hand\nlater). A freeze on\neach agent's assigned subtree turns that from a post-hoc discovery into a\nrefusal at the moment it happens.\n\nUse it whenever: an agent runs unattended, several agents run in parallel,\nor the work touches production or a migration.\n\n**Single writer, even across models.** When more than one model or process\ncontributes to one change, only one of them may write to the filesystem —\nthe rest produce a proposal (a diff, a patch, a plan) for the writer to\napply, never a direct write of their own. Two writers on the same change\nmeans there is no single diff left to review and no way to attribute a\nregression to the process that caused it.\n\n## Gate 4 — Stop-gate: do not declare done before the record is written\n\nRetargeted from a learning-library path convention to this ecosystem's\n`project-memory\u002F`.\n\nStage 6 REMEMBER of `dev-kickoff-edho-ferdian` requires the memory files to\nbe updated \"in the same turn, not later\", and the anti-pattern table\nalready names **Snapshot debt (>3 tasks with no fresh snapshot)** — an\nanticipated failure with no enforcer. This is the enforcer, and it uses\nonly deterministic checks: mtimes and counters, no inference.\n\n| Check | Mechanism | On hit |\n|---|---|---|\n| Task was substantial | count of Edit\u002FWrite calls this session ≥ 3 | classifies the session as complex; below that, nothing else runs |\n| Memory files untouched | mtime of `project-memory\u002F03-progress.md`, `01-decision-register.md`, `04-instincts.md` older than this session's start | **Block** when the session is complex and none were touched |\n| Rationalization language | regex over the transcript tail | **Warning only**, never a block |\n\nThe rationalization patterns worth watching are the exact phrases that\nprecede a Gate 2 or regression-rule violation: *\"skip tests for now\"*,\n*\"pre-existing bug\"*, *\"this was already broken\"*, *\"good enough for now\"*.\nThey warn and never block, because a regex on prose false-positives and a\ngate that blocks on a false positive gets turned off.\n\nNote the honest limitation: this enforces the **habit** of writing the\nrecord, not the **quality** of what was written. That is what the Reflection\nblock and the audit skills are for. Defense in depth, not a replacement.\n\nReport execution state with exactly one of: `inspected` \u002F `changed locally` \u002F\n`verified locally` \u002F `committed` \u002F `pushed` \u002F `blocked`. Never say *fixed*\nuntil the proving command has been rerun and named. Never say *pushed*\nunless the upstream branch actually moved.\n\n## Choosing gates\n\n| Situation | Gates |\n|---|---|\n| Ordinary interactive work | 4 only |\n| Unfamiliar or high-fan-in codebase | 1 + 2 + 4 |\n| Parallel agents, each owning a subtree | 1 + 2 + 3 (per agent) + 4 |\n| Production, migrations, deploys | 2 + 3, always |\n\n## Rules\n\n- Never substitute self-evaluation for a gate. \"Are you sure?\" always\n  returns yes; that is measured, not assumed.\n- Do not gate every Bash call. Routine commands gate once per session,\n  destructive ones every time. Getting this balance wrong is how gates get\n  switched off, and a disabled gate protects nothing.\n- Do not pre-answer a gate's questions from memory to satisfy it. The\n  investigation is the mechanism; a recited answer is theatre.\n- A gate that cannot persist its state should fail **open** — allow the\n  action and warn — rather than loop.\n- These gates never replace REVIEW or VERIFY. They are pre-action and\n  post-session bookends around a loop that still has to run.\n\n## Appendix — implementing these as hooks (Claude Code)\n\nPortable practice first: every gate above is written so a human or an agent\ncan run it by hand in any harness (D-008). This appendix is the optional\nmechanical binding for Claude Code specifically — `PreToolUse` for gates\n1-3, `Stop` for gate 4 — including hookify rule-file syntax\n(`.claude\u002Fhookify.\u003Crule-name>.local.md`) for the pattern-matching gates.\nNothing above depends on this appendix existing.\n\nOne gate ships ready-made: `hooks\u002Fblock-fs-wide-search.js` (Node, no\ndependencies) covers Gate 2 for Bash and PowerShell: it denies root-level\nrecursive searches outright, escalates every watched destructive command to\nan `ask` prompt (never a silent allow, never an outright deny — the user\nmay well want it), and fails open on anything it cannot parse. Install it with\n`npx eef-install --target claude-hooks --only fs-guard`, which copies it to\n`~\u002F.claude\u002Fhooks\u002F` and adds the registration below to\n`~\u002F.claude\u002Fsettings.json` without touching anything else there (never twice,\nand never into a file it cannot parse); `npx eef-install doctor` then\nconfirms it is registered. To do it by hand instead, copy the file and add:\n\n```json\n\"PreToolUse\": [{\n  \"matcher\": \"Bash|PowerShell\",\n  \"hooks\": [{ \"type\": \"command\", \"command\": \"node \\\"\u003Chome>\u002F.claude\u002Fhooks\u002Fblock-fs-wide-search.js\\\"\", \"timeout\": 10 }]\n}]\n```\n\n## Growth path\n\nThis skill is intentionally single-file — four gates read fine as one\ndocument because they share one voice (mechanical, fact-forcing) rather than\nfour distinct lenses. If a fifth gate is added, or any single gate above\ngrows enough to need its own worked examples\u002Fedge-case catalogue (the way\n`click-path-audit-edho-ferdian` split framework-specific tracing detail out),\nsplit that gate into `references\u002F`, following the domain+lens pattern other\nskills in this ecosystem use — do not let this file creep past ~300 lines to\navoid making that call.\n","# Safe Execution — Edho Ferdian Mode\n\n## Why this exists: reasoning gates cannot catch this class\n\nThis ecosystem is already dense with **reasoning gates** — the 8-gate\nReflection block, the Critique-Correction Loop, EDHO SCAN, the Output\nScorecard. Every one of them asks the model about its own work, after the\nwork is done.\n\nThat has a known ceiling. Asking a model \"did you violate anything?\"\nreliably returns \"no\" — self-evaluation shares the biases that produced the\noutput. What changes behaviour is being made to **produce a fact**: asked\nto list every file that imports a module, the model has to actually run\nGrep and Read, and the resulting context changes what it writes next.\nInternal A\u002FB measurements on gated vs ungated agents put the gap at\nroughly +2.25 points on a 10-point rubric across two tasks — a small\nsample, quoted as directional evidence rather than proof.\n\nSo the gates here are deliberately **not** reasoning gates. Each one either\ndemands an artifact or checks a machine-verifiable fact.\n\n## Gate 1 — Fact-forcing, before the first touch of a file\n\nThree stages, in order. Stopping at DENY is what most guards do and it is\nthe least useful part:\n\n```\nDENY  — block the first Edit\u002FWrite\u002FBash attempt against a target\nFORCE — state exactly which facts must be produced\nALLOW — permit the retry once those facts are on the table\n```\n\n**Before editing an existing file**, produce:\n1. Every file that imports or requires it (search the tree — Grep\u002FGlob; if\n   `salak` is on PATH, `salak query importers \u003Cfile>` and, for Go,\n   `salak query callers \u003Cnode>` give the edges with provenance — exit 2\n   means not tracked, so Grep instead; see\n   `dev-kickoff-edho-ferdian\u002Freferences\u002Fsalak-integration.md`).\n2. The public functions or classes this change affects.\n3. If it reads or writes data files: the field names, structure, and date\n   format — using redacted or synthetic values, never raw production data.\n4. The user's current instruction, quoted verbatim.\n\n**Before creating a new file**, produce:\n1. The file(s) and line(s) that will call it.\n2. Confirmation that no existing file already serves that purpose.\n3. Data-shape facts, as above, if it touches data files.\n4. The user's current instruction, quoted verbatim.\n\nFact 3 is the one people skip and the one that pays. In both A\u002FB trials\nthe ungated agent assumed ISO-8601 dates while the real data used\n`%Y\u002F%m\u002F%d %H:%M`. Checking the actual shape kills that entire bug class.\n\nFact 4 looks like ceremony and is not: re-reading the literal instruction\nbefore acting is the cheapest defence against the drift where an agent, six\ntool calls deep, is solving an adjacent problem it invented.\n\n**Scale the facts to the change, not to the file.** Read the edit itself\n(old and new text) before choosing which facts it owes:\n\n| The edit… | Owes |\n|---|---|\n| changes only comments, whitespace or formatting — same code once those are stripped | nothing; proceed |\n| changes a body, but no declaration that other files can see | the call sites that rely on this behaviour (narrower than fact 1), plus fact 4 |\n| touches the public surface — an exported\u002F`pub`\u002Fcapitalised name, a signature, a module's export list, a whole-file Write | fact 1 and fact 2 in full |\n| touches data — a format, schema, query, date\u002Ftime handling, file or network I\u002FO | fact 3, whatever else it owes |\n| cannot be classified with confidence | every fact; uncertainty never earns a discount |\n\nOver-matching is fine (a body inside an exported object counts as surface);\nunder-matching is the failure. This keeps the gate expensive exactly where\nthe A\u002FB bugs lived — data shape and public contracts — and nearly free for\nthe edits that never carried them.\n\nScope: **first touch per target, not every edit.** Once the facts are\npresented, further edits to that same file proceed freely. A gate that\nfires on every call gets disabled by lunchtime.\n\n## Gate 2 — Destructive commands, every single time\n\nUnlike Gate 1, this one never gets a free pass after the first hit.\n\nWatched: `rm -rf` (especially against `\u002F`, `~`, or a project root),\n`git reset --hard`, `git push --force`, `git checkout .`, `git clean -fd`,\n`DROP TABLE` \u002F `DROP DATABASE`, `docker system prune`, `kubectl delete`,\n`chmod 777`, `sudo rm`, `npm publish`, and anything carrying `--no-verify`.\n\nOn a hit, produce before running:\n1. Every file or dataset the command will modify or delete.\n2. A one-line rollback procedure.\n3. The user's current instruction, quoted verbatim.\n\nIf step 2 cannot be written, that is the answer: do not run it.\n\nPowerShell has the same list under other names: `Remove-Item -Recurse`\n(and `rm`\u002F`ri`\u002F`del`\u002F`rd` with `-Recurse` or `\u002Fs`), `Format-Volume`,\n`Clear-Disk`, `Clear-RecycleBin`. In Claude Code, the hook in the appendix\nturns every hit into a permission prompt the user must answer — even in\nauto mode — so the three facts above are what they read before approving.\n\n`--no-verify` is on the list deliberately — it is the mechanical form of\n\"skip the checks\", and it is exactly what an agent under time pressure\nreaches for.\n\n**Runaway searches are destructive too** — of time and of the machine. A\nrecursive search starting at a filesystem, drive, or home root (`find \u002F`,\n`find ~`, `rg x \u002F`, `Get-ChildItem -Recurse C:\\`, `dir \u002Fs C:\\`) never \u003C!-- fs-search-ok: names the banned forms -->\nneeds the three-step justification above: it is simply refused. On Windows\nit runs for hours and outlives the tool call as an orphaned process — this\necosystem's own agent wrappers once spawned six `find.exe` scans hunting for\na `SKILL.md`, some alive for ten hours. Search a known directory instead;\nif the target has no known location, ask. The Claude Code binding is\n`hooks\u002Fblock-fs-wide-search.js` in this skill (see the appendix).\n\nNever build a shell command by interpolating a string. Pass the executable\nand its arguments as separate entries with `shell: false` — string\ninterpolation is how a filename with a space or a quote becomes an\narbitrary command.\n\n## Gate 3 — Freeze: scope the write surface\n\nFreeze Mode locks Write\u002FEdit to a named\nsubtree; reads stay unrestricted. Anything outside is refused with an\nexplanation rather than silently allowed.\n\nThis is the directly load-bearing gate for how this ecosystem actually\nworks. Its own porting effort dispatches several Sonnet agents in parallel,\neach owning a slice of the tree — and the kelompok 3 batch found real\ndamage of exactly this shape afterwards (an agent leaving stale pointers\ninto out-of-scope paths across files outside its brief, fixed by hand\nlater). A freeze on\neach agent's assigned subtree turns that from a post-hoc discovery into a\nrefusal at the moment it happens.\n\nUse it whenever: an agent runs unattended, several agents run in parallel,\nor the work touches production or a migration.\n\n**Single writer, even across models.** When more than one model or process\ncontributes to one change, only one of them may write to the filesystem —\nthe rest produce a proposal (a diff, a patch, a plan) for the writer to\napply, never a direct write of their own. Two writers on the same change\nmeans there is no single diff left to review and no way to attribute a\nregression to the process that caused it.\n\n## Gate 4 — Stop-gate: do not declare done before the record is written\n\nRetargeted from a learning-library path convention to this ecosystem's\n`project-memory\u002F`.\n\nStage 6 REMEMBER of `dev-kickoff-edho-ferdian` requires the memory files to\nbe updated \"in the same turn, not later\", and the anti-pattern table\nalready names **Snapshot debt (>3 tasks with no fresh snapshot)** — an\nanticipated failure with no enforcer. This is the enforcer, and it uses\nonly deterministic checks: mtimes and counters, no inference.\n\n| Check | Mechanism | On hit |\n|---|---|---|\n| Task was substantial | count of Edit\u002FWrite calls this session ≥ 3 | classifies the session as complex; below that, nothing else runs |\n| Memory files untouched | mtime of `project-memory\u002F03-progress.md`, `01-decision-register.md`, `04-instincts.md` older than this session's start | **Block** when the session is complex and none were touched |\n| Rationalization language | regex over the transcript tail | **Warning only**, never a block |\n\nThe rationalization patterns worth watching are the exact phrases that\nprecede a Gate 2 or regression-rule violation: *\"skip tests for now\"*,\n*\"pre-existing bug\"*, *\"this was already broken\"*, *\"good enough for now\"*.\nThey warn and never block, because a regex on prose false-positives and a\ngate that blocks on a false positive gets turned off.\n\nNote the honest limitation: this enforces the **habit** of writing the\nrecord, not the **quality** of what was written. That is what the Reflection\nblock and the audit skills are for. Defense in depth, not a replacement.\n\nReport execution state with exactly one of: `inspected` \u002F `changed locally` \u002F\n`verified locally` \u002F `committed` \u002F `pushed` \u002F `blocked`. Never say *fixed*\nuntil the proving command has been rerun and named. Never say *pushed*\nunless the upstream branch actually moved.\n\n## Choosing gates\n\n| Situation | Gates |\n|---|---|\n| Ordinary interactive work | 4 only |\n| Unfamiliar or high-fan-in codebase | 1 + 2 + 4 |\n| Parallel agents, each owning a subtree | 1 + 2 + 3 (per agent) + 4 |\n| Production, migrations, deploys | 2 + 3, always |\n\n## Rules\n\n- Never substitute self-evaluation for a gate. \"Are you sure?\" always\n  returns yes; that is measured, not assumed.\n- Do not gate every Bash call. Routine commands gate once per session,\n  destructive ones every time. Getting this balance wrong is how gates get\n  switched off, and a disabled gate protects nothing.\n- Do not pre-answer a gate's questions from memory to satisfy it. The\n  investigation is the mechanism; a recited answer is theatre.\n- A gate that cannot persist its state should fail **open** — allow the\n  action and warn — rather than loop.\n- These gates never replace REVIEW or VERIFY. They are pre-action and\n  post-session bookends around a loop that still has to run.\n\n## Appendix — implementing these as hooks (Claude Code)\n\nPortable practice first: every gate above is written so a human or an agent\ncan run it by hand in any harness (D-008). This appendix is the optional\nmechanical binding for Claude Code specifically — `PreToolUse` for gates\n1-3, `Stop` for gate 4 — including hookify rule-file syntax\n(`.claude\u002Fhookify.\u003Crule-name>.local.md`) for the pattern-matching gates.\nNothing above depends on this appendix existing.\n\nOne gate ships ready-made: `hooks\u002Fblock-fs-wide-search.js` (Node, no\ndependencies) covers Gate 2 for Bash and PowerShell: it denies root-level\nrecursive searches outright, escalates every watched destructive command to\nan `ask` prompt (never a silent allow, never an outright deny — the user\nmay well want it), and fails open on anything it cannot parse. Install it with\n`npx eef-install --target claude-hooks --only fs-guard`, which copies it to\n`~\u002F.claude\u002Fhooks\u002F` and adds the registration below to\n`~\u002F.claude\u002Fsettings.json` without touching anything else there (never twice,\nand never into a file it cannot parse); `npx eef-install doctor` then\nconfirms it is registered. To do it by hand instead, copy the file and add:\n\n```json\n\"PreToolUse\": [{\n  \"matcher\": \"Bash|PowerShell\",\n  \"hooks\": [{ \"type\": \"command\", \"command\": \"node \\\"\u003Chome>\u002F.claude\u002Fhooks\u002Fblock-fs-wide-search.js\\\"\", \"timeout\": 10 }]\n}]\n```\n\n## Growth path\n\nThis skill is intentionally single-file — four gates read fine as one\ndocument because they share one voice (mechanical, fact-forcing) rather than\nfour distinct lenses. If a fifth gate is added, or any single gate above\ngrows enough to need its own worked examples\u002Fedge-case catalogue (the way\n`click-path-audit-edho-ferdian` split framework-specific tracing detail out),\nsplit that gate into `references\u002F`, following the domain+lens pattern other\nskills in this ecosystem use — do not let this file creep past ~300 lines to\navoid making that call.\n",{"name":5,"metadata":24,"description":7},{"last_reviewed":25},"2026-10-27","main","533d32ad55c92d40fbabe93cbe21fe048598fe5a",".zcode\u002Fskills\u002Fsafe-execution-edho-ferdian",0,"https:\u002F\u002Fgithub.com\u002Fedhoferdian\u002FEEF","https:\u002F\u002Fgithub.com\u002Fedhoferdian","Elevating ideas through AI, rising from humble beginnings to mastering the frontiers of the web. Every line of code is open-sourced on GitHub, inviting you to f","2026-10-30 03:34:17",{"source":8,"skillName":5,"commands":35,"prompt":45,"zipAvailable":46},[36,39,42],{"tool":37,"command":38},"npx","npx skills add edhoferdian\u002FEEF --skill safe-execution-edho-ferdian",{"tool":40,"command":41},"bunx","bunx skills add edhoferdian\u002FEEF --skill safe-execution-edho-ferdian",{"tool":43,"command":44},"pnpm","pnpm dlx skills add edhoferdian\u002FEEF --skill safe-execution-edho-ferdian","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fedhoferdian\u002FEEF\u002Fblob\u002Fmain\u002F.zcode\u002Fskills\u002Fsafe-execution-edho-ferdian\u002FSKILL.md",true,[48],{"path":49,"resourceType":50,"resourceTypeDesc":51,"sizeBytes":52,"sourceHtmlUrl":53},"hooks\u002Fblock-fs-wide-search.js","SCRIPT","脚本",7899,"https:\u002F\u002Fgithub.com\u002Fedhoferdian\u002FEEF\u002Fblob\u002Fmain\u002F.zcode\u002Fskills\u002Fsafe-execution-edho-ferdian\u002Fhooks\u002Fblock-fs-wide-search.js",[55,62,69,76,83,90],{"skillNo":56,"name":57,"nameZh":6,"description":58,"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},"01M7C9FY2EWE5NB7M3YW0KJX62","video-style-edho-ferdian","Measure a reference video's look and pacing (colour grade as a 3D LUT, cut rhythm, overlays) and reproduce it: grade, cut and composite new footage, verify numerically, hand off an NLE timeline. Use for \"tiru look video ini\", \"samakan grading\", \"bikin video dengan gaya referensi\", \"LUT dari referensi\", \"cut rhythm\".","https:\u002F\u002Fgithub.com\u002Fedhoferdian\u002FEEF\u002Fblob\u002Fmain\u002Fskills\u002Fvideo-style-edho-ferdian\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},"01M7C9FT74P7JXGQ73116B9WS1","test-authoring-edho-ferdian","Write unit and component tests well: React Testing Library, pytest, Go and Vue, plus regression-test patterns. Use for \"tulis test untuk component ini\", \"bagaimana test hook ini\", \"test yang bagus untuk fitur X\", \"tulis test pytest\u002FGo\u002FVue\", or dev-kickoff's TEST stage.","https:\u002F\u002Fgithub.com\u002Fedhoferdian\u002FEEF\u002Fblob\u002Fmain\u002Fskills\u002Ftest-authoring-edho-ferdian\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},"01M7C9FMPE9W76FS7S8ZZ48ARE","system-design-edho-ferdian","Mid-project architecture decisions: ADRs, trade-off analysis, non-functional requirements, scaling tiers and blast radius. Use for \"desain arsitektur\", \"keputusan teknis besar\", \"bikin ADR\", \"trade-off antara X dan Y\", \"microservices atau monolith\", \"apa yang rusak kalau X diubah\", or a capacity question.","https:\u002F\u002Fgithub.com\u002Fedhoferdian\u002FEEF\u002Fblob\u002Fmain\u002Fskills\u002Fsystem-design-edho-ferdian\u002FSKILL.md",[],[],{"skillNo":77,"name":78,"nameZh":6,"description":79,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":80,"categories":81,"keywords":82,"publishedAt":19},"01M7C9FETATAHKG5WKPT4AP1A2","spec-mining-edho-ferdian","Extract behavioural specs from a codebase with none: WHEN\u002FTHEN requirements and invariants, each anchored to the code that enforces it, unclear behaviour recorded as uncertainty. Use for \"ekstrak spec\", \"buat spec dari kode\", \"reverse-engineer the spec\", \"repo ini tidak ada dokumentasi\", or a missing BEHAVIOR_SPEC in dev-kickoff.","https:\u002F\u002Fgithub.com\u002Fedhoferdian\u002FEEF\u002Fblob\u002Fmain\u002Fskills\u002Fspec-mining-edho-ferdian\u002FSKILL.md",[],[],{"skillNo":84,"name":85,"nameZh":6,"description":86,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":87,"categories":88,"keywords":89,"publishedAt":19},"01M7C9F93YCXFCVRDG4S5QFAH4","skill-authoring-edho-ferdian","Create and govern this ecosystem's skills: search before building, a quality bar, compliance checks, promoting repeated principles to rules, and packaging dist\u002F*.skill. Use for \"bikin skill baru\", \"ada skill buat X gak\", \"fork skill ini\", \"skill gue kepake gak sih\", \"package skill ini\", \"buatkan .skill-nya\".","https:\u002F\u002Fgithub.com\u002Fedhoferdian\u002FEEF\u002Fblob\u002Fmain\u002Fskills\u002Fskill-authoring-edho-ferdian\u002FSKILL.md",[],[],{"skillNo":91,"name":92,"nameZh":6,"description":93,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":94,"categories":95,"keywords":96,"publishedAt":19},"01M7C9F4XZJ59R0GS4YEP4E1RD","skill-audit-edho-ferdian","Audit this repo's skills\u002F for staleness, redundancy, broken cross-references and weak descriptions (not the ~\u002F.claude setup: that is config-hygiene). Use for \"audit skill saya\", \"cek skill yang sudah dibuat\", \"ada yang redundan gak\", \"skill mana yang basi\".","https:\u002F\u002Fgithub.com\u002Fedhoferdian\u002FEEF\u002Fblob\u002Fmain\u002Fskills\u002Fskill-audit-edho-ferdian\u002FSKILL.md",[],[]]