@ikon85/agent-workflow-kit 0.25.0 → 0.26.0

package/.agents/skills/setup-workflow/SKILL.md

@@ -26,7 +26,7 @@ This is a prompt-driven skill, not a deterministic script. Explore, present what
 | `## Agent skills` + `## Prod` in CLAUDE.md **and** AGENTS.md | one-line pointers + deploy target (Sections C/G + Write) |
 | `.github/workflows/agent-workflow-kit-update.yml` | optional tested Kit update pull request for GitHub consumers (Section A2) |
 | `docs/agents/census.md` | optional-census choice, paths, state, and safe disable contract seeded from [census.md](./census.md) (Section A3) |
-| `docs/agents/workflow-capabilities.json` + capability assets | consumer-owned Worktree Lifecycle, Memory Lifecycle, and Workflow Advisories choices (Sections A4–A6) |
+| `docs/agents/workflow-capabilities.json` + capability assets | consumer-owned Worktree Lifecycle, Memory Lifecycle, Workflow Advisories, and Safety Guardrails choices (Sections A4–A7) |
 
 ## Idempotency contract — read before writing anything
 
@@ -57,7 +57,7 @@ Read the current state; don't assume. For every target file, read its first line
 - `CONTEXT.md`, `CONTEXT-MAP.md`, `docs/adr/` — domain-doc layout.
 - `docs/agents/`, `docs/agents/skills/`, `docs/conventions/` — prior output of this skill.
 - `docs/agents/census.md`, `.census/profile.json`, `.census/active.json` — an existing census choice or consumer-owned census to adopt.
-- `docs/agents/workflow-capabilities.json`, `.claude/settings.json` — existing capability choices/profile plus Worktree Lifecycle and Workflow Advisories hook wiring to adopt.
+- `docs/agents/workflow-capabilities.json`, `.claude/settings.json`, `git config --get core.hooksPath` — existing capability choices/profile plus Worktree Lifecycle, Workflow Advisories, and Safety Guardrails wiring to adopt.
 - `.memory/active/`, `.memory/archive/`, and `.memory/receipts/` — consumer-owned Memory Lifecycle policies or recovery evidence to adopt without normalization.
 - `gh auth status` (if GitHub) — is `gh` authenticated, and with which scopes?
 
@@ -218,6 +218,31 @@ repository?"* Offer exactly **Yes**, **Later**, and **No**.
 - **Disable** — remove exact kit-owned commands first, then set `enabled:
   false`; preserve every profile value and unknown key.
 
+### 2f. Section A7 — Optional Safety Guardrails
+
+> Safety Guardrails is a counted group of seven independently selectable
+> protections: four agent-tool guards and three repository-security
+> capabilities. Setup stages and validates the chosen profile plus wiring
+> before activation; disabled rows remain exact no-ops.
+
+Read [safety-guardrails.md](./safety-guardrails.md) in full. Reconcile its
+complete `yes / later / no / existing / disable` matrix. If no choice exists,
+ask: *"Should setup configure the portable Safety Guardrails for this
+repository?"* Offer exactly **Yes**, **Later**, and **No**.
+
+- **Yes** — ask which of the seven counted rows to enable, then reconcile only
+  `safetyGuardrails`, exact selected Agent-hook commands, and the selected
+  native Git-hook path. Report `N of 7 enabled`.
+- **Later / No** — record the choice without Agent- or Git-hook wiring.
+- **Existing** — adopt profile and wiring byte-safely; preserve unknown keys
+  and consumer-specific policies.
+- **Disable** — remove only exact kit-owned Agent commands and Git wiring
+  first, then set `enabled: false`.
+
+Testreporter uses the parity-on profile from the seed. Other consumers are
+opt-in/manual: never copy its pnpm, path, shim, surface, or outage decisions
+without repository evidence and user confirmation.
+
 ### 3. Section B — Triage labels
 
 > When `triage` processes an incoming issue it applies labels (or your tracker's equivalent). It needs strings you've actually configured, or it creates duplicates.
@@ -305,10 +330,11 @@ request changes a recorded choice.
 
 For `docs/agents/workflow-capabilities.json`, reconcile only each selected
 capability section. Never replace the whole file: Project Release, Memory
-Lifecycle, Worktree Lifecycle, Workflow Advisories, future sections, and
+Lifecycle, Worktree Lifecycle, Workflow Advisories, Safety Guardrails, future sections, and
 unknown consumer keys share this profile. Apply the transition and hook
-ownership rules from [worktree-lifecycle.md](./worktree-lifecycle.md) and
-[workflow-advisories.md](./workflow-advisories.md) transactionally.
+ownership rules from [worktree-lifecycle.md](./worktree-lifecycle.md),
+[workflow-advisories.md](./workflow-advisories.md), and
+[safety-guardrails.md](./safety-guardrails.md) transactionally.
 
 For Memory Lifecycle, use only the deterministic setup helper from Section A5.
 Do not copy templates with shell commands or edit the capability profile by

package/.agents/skills/setup-workflow/safety-guardrails.md

@@ -0,0 +1,155 @@
+# Safety Guardrails setup contract
+
+Safety Guardrails is a counted, opt-in capability group backed by the
+consumer-owned `docs/agents/workflow-capabilities.json` profile. Its seven
+rows are selected independently: four protect agent tool calls and three
+bootstrap repository security. No repository name, package manager, hook
+surface, threshold, path, or outage policy is inferred from the kit.
+
+## Choice matrix
+
+| State | Setup action |
+|---|---|
+| `missing` | Ask `yes / later / no`; do not infer or write before the answer. |
+| `yes` | Ask which rows to enable, stage profile plus selected wiring, validate, then activate. |
+| `later` | Record the retryable deferral; do not activate any wiring. |
+| `no` | Record the opt-out; do not activate any wiring. |
+| `existing` | Adopt the consumer profile and existing wiring byte-safely. |
+| `disable` | Remove exact kit-owned Agent and Git wiring first, then set `enabled: false`. |
+
+```json safety-guardrails-setup-effects
+[
+  {"state":"missing","choice":null,"operations":[]},
+  {"state":"yes","choice":"yes","operations":["record-choice","stage-enabled-profile","stage-agent-hook-wiring","stage-git-hook-wiring","activate-staged-profile"]},
+  {"state":"later","choice":"later","operations":["record-choice"]},
+  {"state":"no","choice":"no","operations":["record-choice"]},
+  {"state":"existing","choice":"yes","operations":["adopt-existing"]},
+  {"state":"disable","choice":"yes","operations":["remove-agent-hook-wiring","remove-git-hook-wiring","update-profile-disabled"]}
+]
+```
+
+## Counted capability map
+
+Every historical Safety row maps to one profile decision and one shipped
+artifact. Setup reports the result as `N of 7 enabled`, never as one
+all-or-nothing switch.
+
+```json safety-guardrails-capabilities
+[
+  {"id":"secrets","profilePath":"safetyGuardrails.secrets.enabled","artifact":".claude/hooks/block-secrets.py"},
+  {"id":"packageManager","profilePath":"safetyGuardrails.packageManager.enabled","artifact":".claude/hooks/block-npm-install-in-pnpm.py"},
+  {"id":"doubleBackground","profilePath":"safetyGuardrails.doubleBackground.enabled","artifact":".claude/hooks/block-bg-double-background.py"},
+  {"id":"searchShim","profilePath":"safetyGuardrails.searchShim.enabled","artifact":".claude/hooks/grep-shim-guard.py"},
+  {"id":"gitHooks","profilePath":"safetyGuardrails.repositorySecurity.gitHooks.enabled","artifact":"scripts/security/install-git-hooks.mjs"},
+  {"id":"gitleaks","profilePath":"safetyGuardrails.repositorySecurity.gitleaks.enabled","artifact":"scripts/security/ensure-gitleaks.mjs"},
+  {"id":"dependencyAudit","profilePath":"safetyGuardrails.repositorySecurity.dependencyAudit.enabled","artifact":"scripts/security/audit-gate.mjs"}
+]
+```
+
+## Profile shape
+
+Reconcile only `safetyGuardrails`; preserve every sibling section, unknown key,
+and consumer-owned value. The following is the Testreporter parity-on example,
+not a universal default. Other consumers choose each `enabled` value and fill
+only facts proven in their repository.
+
+```json
+{
+  "safetyGuardrails": {
+    "choice": "yes",
+    "enabled": true,
+    "secrets": {
+      "enabled": true,
+      "sensitiveNames": [".env", ".npmrc", "credentials.json"],
+      "sensitivePathFragments": ["secrets/", ".ssh/"],
+      "safeTemplateSuffixes": [".example", ".sample", ".template", ".dist"]
+    },
+    "packageManager": {
+      "enabled": true,
+      "lockfiles": {
+        "pnpm-lock.yaml": "pnpm",
+        "package-lock.json": "npm",
+        "yarn.lock": "yarn",
+        "bun.lockb": "bun"
+      }
+    },
+    "doubleBackground": {"enabled": true, "surfaces": ["claude"]},
+    "searchShim": {
+      "enabled": true,
+      "detected": true,
+      "commandNames": ["grep", "rg"]
+    },
+    "repositorySecurity": {
+      "gitHooks": {"enabled": true, "hooksPath": ".githooks"},
+      "gitleaks": {"enabled": true, "required": false},
+      "dependencyAudit": {
+        "enabled": true,
+        "packageManager": "pnpm",
+        "outagePolicy": "warn",
+        "blockSeverities": ["high", "critical"]
+      }
+    }
+  }
+}
+```
+
+An empty or disabled row is an honest no-op. `searchShim.detected` is a
+consumer fact, not an installation request. `gitleaks.required: false` keeps
+the verified optional provisioner non-blocking when the platform or network is
+unsupported. Dependency-audit outage policy must be explicit.
+
+## Agent-hook ownership
+
+Wire only enabled rows, using the exact kit-owned commands:
+
+- `secrets` — `python3 "$CLAUDE_PROJECT_DIR/.claude/hooks/block-secrets.py"`
+- `packageManager` —
+  `python3 "$CLAUDE_PROJECT_DIR/.claude/hooks/block-npm-install-in-pnpm.py"`
+- `doubleBackground` —
+  `python3 "$CLAUDE_PROJECT_DIR/.claude/hooks/block-bg-double-background.py"`
+- `searchShim` —
+  `python3 "$CLAUDE_PROJECT_DIR/.claude/hooks/grep-shim-guard.py"`
+
+All four are PreToolUse adapters. Preserve unrelated matchers and commands in
+`.claude/settings.json`. Disable removes only these exact commands.
+
+## Repository wiring
+
+When `gitHooks.enabled` is true, run:
+
+```bash
+node scripts/security/install-git-hooks.mjs <hooksPath>
+```
+
+only after confirming the configured directory exists and contains the
+consumer's intended hooks. Never create or replace consumer hook bodies.
+
+When `gitleaks.enabled` is true, the pinned profile may be provisioned through
+`scripts/security/ensure-gitleaks.mjs`; checksum mismatch, unsupported platform,
+offline state, and unwritable destination remain distinct visible outcomes.
+Never read a secret file to test the provisioner.
+
+When `dependencyAudit.enabled` is true, use
+`scripts/security/audit-gate.mjs` with the configured package manager and
+outage policy. High/Critical findings block; Low/Moderate findings pass.
+
+## Transaction and failure contract
+
+Prepare the next profile and `.claude/settings.json` in same-directory staging
+files. Validate JSON, all selected artifact paths, the seven-row count, the
+configured Git-hook directory, and package-manager facts before mutation.
+Then:
+
+1. reconcile selected Agent commands in the staged settings;
+2. remember the prior `core.hooksPath` and apply the selected Git wiring;
+3. atomically rename staged settings and profile into place;
+4. on any failure, restore the prior Git config and leave active files
+   byte-for-byte unchanged.
+
+Disable reverses the safety order: remove exact Agent commands and restore or
+unset only the kit-owned Git wiring before committing `enabled: false`.
+Repeated activation with the same choices is byte-idempotent.
+
+Setup and verification may inspect paths, profile metadata, command exit
+status, advisory counts, and hashes. They must never read secret contents or
+print environment/config values.

package/.claude/skills/setup-workflow/SKILL.md

@@ -26,7 +26,7 @@ This is a prompt-driven skill, not a deterministic script. Explore, present what
 | `## Agent skills` + `## Prod` in CLAUDE.md **and** AGENTS.md | one-line pointers + deploy target (Sections C/G + Write) |
 | `.github/workflows/agent-workflow-kit-update.yml` | optional tested Kit update pull request for GitHub consumers (Section A2) |
 | `docs/agents/census.md` | optional-census choice, paths, state, and safe disable contract seeded from [census.md](./census.md) (Section A3) |
-| `docs/agents/workflow-capabilities.json` + capability assets | consumer-owned Worktree Lifecycle, Memory Lifecycle, and Workflow Advisories choices (Sections A4–A6) |
+| `docs/agents/workflow-capabilities.json` + capability assets | consumer-owned Worktree Lifecycle, Memory Lifecycle, Workflow Advisories, and Safety Guardrails choices (Sections A4–A7) |
 
 ## Idempotency contract — read before writing anything
 
@@ -57,7 +57,7 @@ Read the current state; don't assume. For every target file, read its first line
 - `CONTEXT.md`, `CONTEXT-MAP.md`, `docs/adr/` — domain-doc layout.
 - `docs/agents/`, `docs/agents/skills/`, `docs/conventions/` — prior output of this skill.
 - `docs/agents/census.md`, `.census/profile.json`, `.census/active.json` — an existing census choice or consumer-owned census to adopt.
-- `docs/agents/workflow-capabilities.json`, `.claude/settings.json` — existing capability choices/profile plus Worktree Lifecycle and Workflow Advisories hook wiring to adopt.
+- `docs/agents/workflow-capabilities.json`, `.claude/settings.json`, `git config --get core.hooksPath` — existing capability choices/profile plus Worktree Lifecycle, Workflow Advisories, and Safety Guardrails wiring to adopt.
 - `.memory/active/`, `.memory/archive/`, and `.memory/receipts/` — consumer-owned Memory Lifecycle policies or recovery evidence to adopt without normalization.
 - `gh auth status` (if GitHub) — is `gh` authenticated, and with which scopes?
 
@@ -218,6 +218,31 @@ repository?"* Offer exactly **Yes**, **Later**, and **No**.
 - **Disable** — remove exact kit-owned commands first, then set `enabled:
   false`; preserve every profile value and unknown key.
 
+### 2f. Section A7 — Optional Safety Guardrails
+
+> Safety Guardrails is a counted group of seven independently selectable
+> protections: four agent-tool guards and three repository-security
+> capabilities. Setup stages and validates the chosen profile plus wiring
+> before activation; disabled rows remain exact no-ops.
+
+Read [safety-guardrails.md](./safety-guardrails.md) in full. Reconcile its
+complete `yes / later / no / existing / disable` matrix. If no choice exists,
+ask: *"Should setup configure the portable Safety Guardrails for this
+repository?"* Offer exactly **Yes**, **Later**, and **No**.
+
+- **Yes** — ask which of the seven counted rows to enable, then reconcile only
+  `safetyGuardrails`, exact selected Agent-hook commands, and the selected
+  native Git-hook path. Report `N of 7 enabled`.
+- **Later / No** — record the choice without Agent- or Git-hook wiring.
+- **Existing** — adopt profile and wiring byte-safely; preserve unknown keys
+  and consumer-specific policies.
+- **Disable** — remove only exact kit-owned Agent commands and Git wiring
+  first, then set `enabled: false`.
+
+Testreporter uses the parity-on profile from the seed. Other consumers are
+opt-in/manual: never copy its pnpm, path, shim, surface, or outage decisions
+without repository evidence and user confirmation.
+
 ### 3. Section B — Triage labels
 
 > When `triage` processes an incoming issue it applies labels (or your tracker's equivalent). It needs strings you've actually configured, or it creates duplicates.
@@ -305,10 +330,11 @@ request changes a recorded choice.
 
 For `docs/agents/workflow-capabilities.json`, reconcile only each selected
 capability section. Never replace the whole file: Project Release, Memory
-Lifecycle, Worktree Lifecycle, Workflow Advisories, future sections, and
+Lifecycle, Worktree Lifecycle, Workflow Advisories, Safety Guardrails, future sections, and
 unknown consumer keys share this profile. Apply the transition and hook
-ownership rules from [worktree-lifecycle.md](./worktree-lifecycle.md) and
-[workflow-advisories.md](./workflow-advisories.md) transactionally.
+ownership rules from [worktree-lifecycle.md](./worktree-lifecycle.md),
+[workflow-advisories.md](./workflow-advisories.md), and
+[safety-guardrails.md](./safety-guardrails.md) transactionally.
 
 For Memory Lifecycle, use only the deterministic setup helper from Section A5.
 Do not copy templates with shell commands or edit the capability profile by

package/.claude/skills/setup-workflow/safety-guardrails.md

@@ -0,0 +1,155 @@
+# Safety Guardrails setup contract
+
+Safety Guardrails is a counted, opt-in capability group backed by the
+consumer-owned `docs/agents/workflow-capabilities.json` profile. Its seven
+rows are selected independently: four protect agent tool calls and three
+bootstrap repository security. No repository name, package manager, hook
+surface, threshold, path, or outage policy is inferred from the kit.
+
+## Choice matrix
+
+| State | Setup action |
+|---|---|
+| `missing` | Ask `yes / later / no`; do not infer or write before the answer. |
+| `yes` | Ask which rows to enable, stage profile plus selected wiring, validate, then activate. |
+| `later` | Record the retryable deferral; do not activate any wiring. |
+| `no` | Record the opt-out; do not activate any wiring. |
+| `existing` | Adopt the consumer profile and existing wiring byte-safely. |
+| `disable` | Remove exact kit-owned Agent and Git wiring first, then set `enabled: false`. |
+
+```json safety-guardrails-setup-effects
+[
+  {"state":"missing","choice":null,"operations":[]},
+  {"state":"yes","choice":"yes","operations":["record-choice","stage-enabled-profile","stage-agent-hook-wiring","stage-git-hook-wiring","activate-staged-profile"]},
+  {"state":"later","choice":"later","operations":["record-choice"]},
+  {"state":"no","choice":"no","operations":["record-choice"]},
+  {"state":"existing","choice":"yes","operations":["adopt-existing"]},
+  {"state":"disable","choice":"yes","operations":["remove-agent-hook-wiring","remove-git-hook-wiring","update-profile-disabled"]}
+]
+```
+
+## Counted capability map
+
+Every historical Safety row maps to one profile decision and one shipped
+artifact. Setup reports the result as `N of 7 enabled`, never as one
+all-or-nothing switch.
+
+```json safety-guardrails-capabilities
+[
+  {"id":"secrets","profilePath":"safetyGuardrails.secrets.enabled","artifact":".claude/hooks/block-secrets.py"},
+  {"id":"packageManager","profilePath":"safetyGuardrails.packageManager.enabled","artifact":".claude/hooks/block-npm-install-in-pnpm.py"},
+  {"id":"doubleBackground","profilePath":"safetyGuardrails.doubleBackground.enabled","artifact":".claude/hooks/block-bg-double-background.py"},
+  {"id":"searchShim","profilePath":"safetyGuardrails.searchShim.enabled","artifact":".claude/hooks/grep-shim-guard.py"},
+  {"id":"gitHooks","profilePath":"safetyGuardrails.repositorySecurity.gitHooks.enabled","artifact":"scripts/security/install-git-hooks.mjs"},
+  {"id":"gitleaks","profilePath":"safetyGuardrails.repositorySecurity.gitleaks.enabled","artifact":"scripts/security/ensure-gitleaks.mjs"},
+  {"id":"dependencyAudit","profilePath":"safetyGuardrails.repositorySecurity.dependencyAudit.enabled","artifact":"scripts/security/audit-gate.mjs"}
+]
+```
+
+## Profile shape
+
+Reconcile only `safetyGuardrails`; preserve every sibling section, unknown key,
+and consumer-owned value. The following is the Testreporter parity-on example,
+not a universal default. Other consumers choose each `enabled` value and fill
+only facts proven in their repository.
+
+```json
+{
+  "safetyGuardrails": {
+    "choice": "yes",
+    "enabled": true,
+    "secrets": {
+      "enabled": true,
+      "sensitiveNames": [".env", ".npmrc", "credentials.json"],
+      "sensitivePathFragments": ["secrets/", ".ssh/"],
+      "safeTemplateSuffixes": [".example", ".sample", ".template", ".dist"]
+    },
+    "packageManager": {
+      "enabled": true,
+      "lockfiles": {
+        "pnpm-lock.yaml": "pnpm",
+        "package-lock.json": "npm",
+        "yarn.lock": "yarn",
+        "bun.lockb": "bun"
+      }
+    },
+    "doubleBackground": {"enabled": true, "surfaces": ["claude"]},
+    "searchShim": {
+      "enabled": true,
+      "detected": true,
+      "commandNames": ["grep", "rg"]
+    },
+    "repositorySecurity": {
+      "gitHooks": {"enabled": true, "hooksPath": ".githooks"},
+      "gitleaks": {"enabled": true, "required": false},
+      "dependencyAudit": {
+        "enabled": true,
+        "packageManager": "pnpm",
+        "outagePolicy": "warn",
+        "blockSeverities": ["high", "critical"]
+      }
+    }
+  }
+}
+```
+
+An empty or disabled row is an honest no-op. `searchShim.detected` is a
+consumer fact, not an installation request. `gitleaks.required: false` keeps
+the verified optional provisioner non-blocking when the platform or network is
+unsupported. Dependency-audit outage policy must be explicit.
+
+## Agent-hook ownership
+
+Wire only enabled rows, using the exact kit-owned commands:
+
+- `secrets` — `python3 "$CLAUDE_PROJECT_DIR/.claude/hooks/block-secrets.py"`
+- `packageManager` —
+  `python3 "$CLAUDE_PROJECT_DIR/.claude/hooks/block-npm-install-in-pnpm.py"`
+- `doubleBackground` —
+  `python3 "$CLAUDE_PROJECT_DIR/.claude/hooks/block-bg-double-background.py"`
+- `searchShim` —
+  `python3 "$CLAUDE_PROJECT_DIR/.claude/hooks/grep-shim-guard.py"`
+
+All four are PreToolUse adapters. Preserve unrelated matchers and commands in
+`.claude/settings.json`. Disable removes only these exact commands.
+
+## Repository wiring
+
+When `gitHooks.enabled` is true, run:
+
+```bash
+node scripts/security/install-git-hooks.mjs <hooksPath>
+```
+
+only after confirming the configured directory exists and contains the
+consumer's intended hooks. Never create or replace consumer hook bodies.
+
+When `gitleaks.enabled` is true, the pinned profile may be provisioned through
+`scripts/security/ensure-gitleaks.mjs`; checksum mismatch, unsupported platform,
+offline state, and unwritable destination remain distinct visible outcomes.
+Never read a secret file to test the provisioner.
+
+When `dependencyAudit.enabled` is true, use
+`scripts/security/audit-gate.mjs` with the configured package manager and
+outage policy. High/Critical findings block; Low/Moderate findings pass.
+
+## Transaction and failure contract
+
+Prepare the next profile and `.claude/settings.json` in same-directory staging
+files. Validate JSON, all selected artifact paths, the seven-row count, the
+configured Git-hook directory, and package-manager facts before mutation.
+Then:
+
+1. reconcile selected Agent commands in the staged settings;
+2. remember the prior `core.hooksPath` and apply the selected Git wiring;
+3. atomically rename staged settings and profile into place;
+4. on any failure, restore the prior Git config and leave active files
+   byte-for-byte unchanged.
+
+Disable reverses the safety order: remove exact Agent commands and restore or
+unset only the kit-owned Git wiring before committing `enabled: false`.
+Repeated activation with the same choices is byte-idempotent.
+
+Setup and verification may inspect paths, profile metadata, command exit
+status, advisory counts, and hashes. They must never read secret contents or
+print environment/config values.

package/README.md

@@ -332,6 +332,24 @@ still reference. Flags: `--force` (overwrite pre-existing files on `init`),
 
 ## Release notes
 
+### 0.26.0
+
+- added: `.agents/skills/setup-workflow/safety-guardrails.md`
+- added: `.claude/hooks/_safety_guard.py`
+- added: `.claude/hooks/block-bg-double-background.py`
+- added: `.claude/hooks/block-npm-install-in-pnpm.py`
+- added: `.claude/hooks/block-secrets.py`
+- added: `.claude/hooks/grep-shim-guard.py`
+- added: `.claude/skills/setup-workflow/safety-guardrails.md`
+- added: `scripts/safety-guardrails/core.py`
+- added: `scripts/safety-guardrails/search.py`
+- added: `scripts/security/audit-gate.mjs`
+- added: `scripts/security/ensure-gitleaks.mjs`
+- added: `scripts/security/gitleaks-profile.json`
+- added: `scripts/security/install-git-hooks.mjs`
+- changed: `.agents/skills/setup-workflow/SKILL.md`
+- changed: `.claude/skills/setup-workflow/SKILL.md`
+
 ### 0.25.0
 
 - added: `.claude/hooks/convention-drift-hint.py`

package/agent-workflow-kit.package.json

@@ -1,5 +1,5 @@
 {
-  "kitVersion": "0.25.0",
+  "kitVersion": "0.26.0",
   "files": [
     {
       "path": ".agents/skills/ask-matt/SKILL.md",
@@ -568,12 +568,21 @@
       "mode": 420,
       "origin": "kit"
     },
+    {
+      "path": ".agents/skills/setup-workflow/safety-guardrails.md",
+      "kind": "skill",
+      "ownerSkill": "setup-workflow",
+      "surface": "codex",
+      "sha256": "08793fc7c9164cde9f7e09154e02d5718428272b85fb67d74681d3a415d4d483",
+      "mode": 420,
+      "origin": "kit"
+    },
     {
       "path": ".agents/skills/setup-workflow/SKILL.md",
       "kind": "skill",
       "ownerSkill": "setup-workflow",
       "surface": "codex",
-      "sha256": "96c0c3ba9ed6b194030a8c45be618caf934e9c4ca8415ea19d726af33cd763cb",
+      "sha256": "b012c477612944d276e7012461f3ec38ecc25b5d92966b667150e1fb547b20cb",
       "mode": 420,
       "origin": "kit"
     },
@@ -890,6 +899,13 @@
       "mode": 420,
       "origin": "kit"
     },
+    {
+      "path": ".claude/hooks/_safety_guard.py",
+      "kind": "hook",
+      "sha256": "1f5c387b11731c7b1cd027fc58acd4c1e45c64799975dc50eeff72469df0deb0",
+      "mode": 420,
+      "origin": "kit"
+    },
     {
       "path": ".claude/hooks/baseline-capture-hint.py",
       "kind": "hook",
@@ -897,6 +913,27 @@
       "mode": 493,
       "origin": "kit"
     },
+    {
+      "path": ".claude/hooks/block-bg-double-background.py",
+      "kind": "hook",
+      "sha256": "ad0c763e3afbb25685f91eea2c39fc3ee58c765536a961810a49bd0c7bbd47b8",
+      "mode": 493,
+      "origin": "kit"
+    },
+    {
+      "path": ".claude/hooks/block-npm-install-in-pnpm.py",
+      "kind": "hook",
+      "sha256": "54da075f8e63e5e927cb752c26c47880da49eb1f730f078558f3f5eec9377c82",
+      "mode": 493,
+      "origin": "kit"
+    },
+    {
+      "path": ".claude/hooks/block-secrets.py",
+      "kind": "hook",
+      "sha256": "962c4f7f035c5cb4c322ee7c76b707a148526000427f4ebd49bfcac50af003b4",
+      "mode": 493,
+      "origin": "kit"
+    },
     {
       "path": ".claude/hooks/branch-context.py",
       "kind": "hook",
@@ -946,6 +983,13 @@
       "mode": 493,
       "origin": "kit"
     },
+    {
+      "path": ".claude/hooks/grep-shim-guard.py",
+      "kind": "hook",
+      "sha256": "9f69541c60c795bd5ec2a4f5d155f6fd3ced4f2ac6a28ed0c55c9a57772dd0d2",
+      "mode": 493,
+      "origin": "kit"
+    },
     {
       "path": ".claude/hooks/loc-offender-forewarn.py",
       "kind": "hook",
@@ -1702,12 +1746,21 @@
       "mode": 420,
       "origin": "kit"
     },
+    {
+      "path": ".claude/skills/setup-workflow/safety-guardrails.md",
+      "kind": "skill",
+      "ownerSkill": "setup-workflow",
+      "surface": "claude",
+      "sha256": "08793fc7c9164cde9f7e09154e02d5718428272b85fb67d74681d3a415d4d483",
+      "mode": 420,
+      "origin": "kit"
+    },
     {
       "path": ".claude/skills/setup-workflow/SKILL.md",
       "kind": "skill",
       "ownerSkill": "setup-workflow",
       "surface": "claude",
-      "sha256": "96c0c3ba9ed6b194030a8c45be618caf934e9c4ca8415ea19d726af33cd763cb",
+      "sha256": "b012c477612944d276e7012461f3ec38ecc25b5d92966b667150e1fb547b20cb",
       "mode": 420,
       "origin": "kit"
     },
@@ -2268,6 +2321,48 @@
       "mode": 420,
       "origin": "kit"
     },
+    {
+      "path": "scripts/safety-guardrails/core.py",
+      "kind": "script",
+      "sha256": "c69c9d4e1b17efca51de12454b119e9ee6758aa6c8d73fe64e4bafd76c804f4e",
+      "mode": 420,
+      "origin": "kit"
+    },
+    {
+      "path": "scripts/safety-guardrails/search.py",
+      "kind": "script",
+      "sha256": "345b1456f1a00320b6ad3e4fad1b050d8dd21a16b51b5e5cfdaf6a579530eac1",
+      "mode": 420,
+      "origin": "kit"
+    },
+    {
+      "path": "scripts/security/audit-gate.mjs",
+      "kind": "script",
+      "sha256": "c6129e77f58e80f5880ebd6d2d235e27e9e31a20ee3cd967ce64815738e61e54",
+      "mode": 493,
+      "origin": "kit"
+    },
+    {
+      "path": "scripts/security/ensure-gitleaks.mjs",
+      "kind": "script",
+      "sha256": "5591f31d88362aa0df75f7b6bb7f22c094822b4bd00c5f6bf9ccff6a2313ed56",
+      "mode": 493,
+      "origin": "kit"
+    },
+    {
+      "path": "scripts/security/gitleaks-profile.json",
+      "kind": "doc",
+      "sha256": "2480f7b4656d28cea1ede895970275d7b97f58d396341ee5e36b3df12575c1b6",
+      "mode": 420,
+      "origin": "kit"
+    },
+    {
+      "path": "scripts/security/install-git-hooks.mjs",
+      "kind": "script",
+      "sha256": "8cf43120ea485fae7dc6aae38a14a3e703dbd18422ef13a199aedb3ec66b0184",
+      "mode": 493,
+      "origin": "kit"
+    },
     {
       "path": "scripts/workflow-advisories/capabilities.json",
       "kind": "doc",

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@ikon85/agent-workflow-kit",
-  "version": "0.25.0",
+  "version": "0.26.0",
   "description": "Portable AI-agent workflow skills (plan → execute → land → learn) for Claude Code & Codex — grilling, TDD, diagnosis, two-axis code review, cross-model Codex review, design & domain-modeling, plus a skill router (ask-matt). npx init/update/diff/uninstall.",
   "type": "module",
   "bin": {

package/scripts/test_skill_publish_audit.py

@@ -124,6 +124,13 @@ def _known_hashes(root: Path) -> set:
         for f in data.get("files", []):
             if "sha256" in f:
                 hashes.add(f["sha256"])
+    gitleaks = root / "scripts/security/gitleaks-profile.json"
+    if gitleaks.exists():
+        data = json.loads(gitleaks.read_text(encoding="utf-8"))
+        for platform in data.get("platforms", {}).values():
+            checksum = platform.get("sha256")
+            if isinstance(checksum, str):
+                hashes.add(checksum)
     return hashes
 
 
@@ -293,6 +300,14 @@ class Exemptions(unittest.TestCase):
                 {"path": "x", "sha256": "f" * 64}]}), encoding="utf-8")
         self.assertEqual(audit_dir(self.dir), [])
 
+    def test_gitleaks_profile_checksums_are_verified_pin_metadata(self):
+        profile = self.dir / "scripts/security"
+        profile.mkdir(parents=True)
+        (profile / "gitleaks-profile.json").write_text(json.dumps({
+            "platforms": {"linux-x64": {"sha256": "e" * 64}},
+        }), encoding="utf-8")
+        self.assertEqual(audit_dir(self.dir), [])
+
     def test_loc_offender_traversal_guard_allowed(self):
         p = self.dir / "scripts"
         p.mkdir()

package/src/lib/bundle.mjs

@@ -104,6 +104,20 @@ export const HELPER_FILES = [
   { path: '.claude/hooks/migration-snapshot-reminder.py', kind: 'hook', mode: 0o755 },
   { path: '.claude/hooks/loc-offender-forewarn.py', kind: 'hook', mode: 0o755 },
   { path: '.claude/hooks/drift-guard.py', kind: 'hook', mode: 0o755 },
+  // Counted Safety Guardrails unit: shared policy/search core, one loader,
+  // four thin Agent adapters, and three portable repository-security
+  // primitives. Activation stays consumer-owned through setup-workflow.
+  { path: 'scripts/safety-guardrails/core.py', kind: 'script', mode: 0o644 },
+  { path: 'scripts/safety-guardrails/search.py', kind: 'script', mode: 0o644 },
+  { path: '.claude/hooks/_safety_guard.py', kind: 'hook', mode: 0o644 },
+  { path: '.claude/hooks/block-secrets.py', kind: 'hook', mode: 0o755 },
+  { path: '.claude/hooks/block-npm-install-in-pnpm.py', kind: 'hook', mode: 0o755 },
+  { path: '.claude/hooks/block-bg-double-background.py', kind: 'hook', mode: 0o755 },
+  { path: '.claude/hooks/grep-shim-guard.py', kind: 'hook', mode: 0o755 },
+  { path: 'scripts/security/install-git-hooks.mjs', kind: 'script', mode: 0o755 },
+  { path: 'scripts/security/ensure-gitleaks.mjs', kind: 'script', mode: 0o755 },
+  { path: 'scripts/security/gitleaks-profile.json', kind: 'doc', mode: 0o644 },
+  { path: 'scripts/security/audit-gate.mjs', kind: 'script', mode: 0o755 },
   // SessionStart skill-freshness drift-hint (audit-skills names it). For each
   // <skill>/SOURCES.txt it flags sources newer in git than the SKILL.md. Imports
   // _hook_utils (shipped above); stdlib-only otherwise. Executable hook → 0o755.