gitar-bot · Oct 17, 2026 · Oct 17, 2026 · Oct 17, 2026 · Oct 17, 2026 · Oct 17, 2026
@@ -1,11 +1,13 @@
 ---
 name: prompt-perfector
-description: Refine or evaluate LLM and agent prompts while preserving intent with explicit output and action controls. Use when asked to polish, perfect, rewrite, structure, optimize, or assess a prompt.
+description: Refine or evaluate prompts, preserving intent, output controls and authority. Use for prompt refinement or assessment.
 ---
 
 # Prompt Perfector
 
-Produce a ready-to-use prompt that preserves intent. Refine only unless evaluation or execution is explicit.
+Refine only unless evaluation or execution is explicit.
+
+Claude’s separately named `prompt` skill refines then executes within existing authority. These are distinct defaults; explicit user controls win.
 
 ## Workflow
 

@@ -22,7 +22,7 @@ Follow `AGENTS.md` review throttling and `docs/codex-review-protocol.md` before
 
 ## Gate
 
-For UI/browser/styling/routing/reduced-motion/forced-colors changes, run `npm run ensure` before browser QA (never assume port 3000/3001/3002), then use `npm run verify:ui` (Chromium) as the gate. These are local/offline-safe (demo mode).
+Follow `docs/agents/verification-gates.md#gate-selection`. For UI/browser/styling/routing/reduced-motion/forced-colors changes, run `npm run ensure` before browser QA, use its printed URL and verify `/api/local-project-id`, then prove the changed owner or journey first. Use `npm run verify:ui` for shared UI foundations or when PR/handoff policy requires the complete Chromium gate. Report focused proof as focused proof, retaining required broad gates and any physical Safari or installed-PWA acceptance gaps.
 
 ## Review Checklist
 

@@ -23,7 +23,7 @@ Follow `AGENTS.md` review throttling and `docs/codex-review-protocol.md` before
 
 ## Provider boundary
 
-Ingestion checks against live services and any reindex that spends OpenAI budget are confirmation-required (`AGENTS.md`). Report the command and ask. Note the worker stack (Deno edge functions, Python OCR in `worker/python/requirements.txt`) does not run in this VM — do not attempt to execute it.
+Ingestion checks against live services and any reindex that spends OpenAI budget are confirmation-required (`AGENTS.md`). Report the command and ask. Inspect current-host capabilities before selecting offline worker checks: confirm the required Deno/Python binaries and the specific command's local dependencies, fixtures and service needs. Run supported offline checks when available; report missing capabilities and unrun evidence precisely. A provisioned runtime does not authorize live ingestion, reindexing, downloads or provider spend.
 
 ## Review Checklist
 

@@ -23,7 +23,7 @@ Default to running **only** offline-safe gates. For any provider-touching comman
 
 - `lint`, `typecheck`, `test`
 - `verify:cheap` (lock parity + diff-integrity + lint + typecheck + unit tests — nothing else)
-- `verify:full` (41 static/consistency gates, then lint + typecheck + test; does not invoke cheap)
+- `verify:full` (static/consistency gates, then lint + typecheck + test; does not invoke cheap)
 - `verify:pr-local` (risk-routed PR mirror: focused docs/workflow contracts for light scope; fail-closed lint/typecheck/full unit plus applicable build/domain checks for executable or unknown scope); inspect selection with `verify:pr-local -- --dry-run --files <paths>`
 - `eval:rag:offline`, `ensure`, `verify:ui` (Chromium, local dev server / demo mode)
 
@@ -37,9 +37,9 @@ Default to running **only** offline-safe gates. For any provider-touching comman
 ## Routing by touched surface
 
 - **Retrieval / ranking / selection / chunking / scoring:** offline gate is `verify:cheap`; note that the merge gate `eval:retrieval:quality` (36/36) is provider-touching → report and ask.
-- **Ingestion / answer-gen / source-governance / privacy / production env:** smallest relevant domain check + `check:production-readiness` (runs fail-closed offline).
+- **Ingestion / answer-gen / source-governance / privacy / production env:** follow `docs/agents/verification-gates.md#gate-selection`: smallest relevant offline domain check; add `check:production-readiness:ci` when readiness or configuration structure is in scope. The bare `check:production-readiness` is release-mode proof for production release or activation, retaining all applicable approvals.
 - **UI / frontend / a11y / routing / styling:** `npm run ensure`, then the affected journey;
-  add `verify:ui` only for shared UI foundations or a distinct unclosed failure class.
+  follow `docs/agents/verification-gates.md#gate-selection` and add `verify:ui` for shared UI foundations or when PR/handoff policy requires the complete Chromium gate.
 - **Supabase env/config change:** `check:supabase-project` (provider — report and ask).
 - **Default source/config/test change:** `verify:pr-local` at PR handoff; do not pre-run
   `verify:cheap` because the routed handoff gate already owns that coverage.

@@ -7,6 +7,8 @@ description: Perfect a rough request or draft prompt using relevant context, att
 
 Turn the user's request and relevant supplied context into a precise, ready-to-use prompt, then carry it out. Preserve intent, avoid scope creep, and use only the detail, context, tools, and validation that change the result.
 
+**Default for this skill:** refine, show the perfected prompt, then execute within existing authority. The separately named repository `prompt-perfector` skill refines only unless execution is explicit; it is not an alias. Preserve the invoked skill's default and the user's latest explicit controls.
+
 ## Operating principles
 
 - Put the outcome, constraints, and success criteria ahead of process detail.

@@ -1,27 +1,26 @@
 ---
 name: semgrep
 description: >-
-  Run Semgrep static analysis scan on a codebase using parallel subagents.
+  Run Semgrep static analysis on a codebase with capability-checked workers.
   Supports two scan modes — "run all" (full ruleset coverage) and "important
-  only" (high-confidence security vulnerabilities). Automatically detects and
-  uses Semgrep Pro for cross-file taint analysis when available. Use when asked
+  only" (high-confidence security vulnerabilities). Records local Semgrep Pro
+  evidence and uses the approved engine. Use when asked
   to scan code for vulnerabilities, run a security audit with Semgrep, find
-  bugs, or perform static analysis. Spawns parallel workers for multi-language
-  codebases.
-allowed-tools: Bash Read Glob Task AskUserQuestion TaskCreate TaskList TaskUpdate
+  bugs, or perform static analysis. Uses parallel workers when authorized and
+  supported, with bounded direct execution as a fallback.
 ---
 
 # Semgrep Security Scan
 
-Run a Semgrep scan with automatic language detection, parallel execution via Task subagents, and merged SARIF output.
+Run a Semgrep scan with local language detection, supported execution tools, and merged SARIF output. Inspect current-host capabilities; this seed does not install a scanner agent or prove hosted tool availability.
 
 ## Essential Principles
 
-1. **Always use `--metrics=off`** — Semgrep sends telemetry by default; `--config auto` also phones home. Every `semgrep` command must include `--metrics=off` to prevent data leakage during security audits.
-2. **User must approve the scan plan (Step 3 is a hard gate)** — The original "scan this codebase" request is NOT approval. Present exact rulesets, target, engine, and mode; wait for explicit "yes"/"proceed" before spawning scanners.
+1. **Always use `--metrics=off`** — Semgrep sends telemetry by default; `--config auto` also phones home. Every `semgrep` command must include `--metrics=off`. This disables telemetry, not registry or provider access; those actions still need authorization.
+2. **User must approve the scan plan (Step 3 is a hard gate)** — A generic "scan this codebase" request does not approve an unspecified plan. Present exact rulesets, target, engine, mode and network/provider actions before execution. Preserve existing explicit authorization for that exact plan; ask only for missing authority.
 3. **Third-party rulesets are required, not optional** — Trail of Bits, 0xdea, and Decurity rules catch vulnerabilities absent from the official registry. Include them whenever the detected language matches.
-4. **Spawn all scan Tasks in a single message** — Parallel execution is the core performance advantage. Never spawn Tasks sequentially; always emit all Task tool calls in one response.
-5. **Always check for Semgrep Pro before scanning** — Pro enables cross-file taint tracking and catches ~250% more true positives. Skipping the check means silently missing critical inter-file vulnerabilities.
+4. **Use supported execution capabilities** — When delegation is authorized and workers are available, dispatch independent language tasks together using the bounded scanner prompt. Otherwise execute the same approved tasks directly within current-host limits. Do not invent an agent type, tool or delegation authority.
+5. **Record engine evidence honestly** — Inspect only local binary/version/installation metadata before approval. Pro supports cross-file analysis; installed metadata does not prove a usable license. Any Pro, registry or network validation belongs after Step 3 authorization. Stop on an unavailable approved engine; do not silently substitute OSS.
 
 ## When to Use
 
@@ -79,15 +78,12 @@ $OUTPUT_DIR/
 
 ## Prerequisites
 
-**Required:** Semgrep CLI (`semgrep --version`). If not installed, see [Semgrep installation docs](https://semgrep.dev/docs/getting-started/).
+**Required:** Semgrep CLI. Discover it locally with `command -v semgrep` and `SEMGREP_ENABLE_VERSION_CHECK=0 semgrep --metrics=off --version`. If absent, report the missing capability; installation needs applicable authority. See [Semgrep installation docs](https://semgrep.dev/docs/getting-started/).
 
-**Optional:** Semgrep Pro — enables cross-file taint tracking, inter-procedural analysis, and additional languages (Apex, C#, Elixir). Check with:
-
-```bash
-semgrep --pro --validate --config p/default 2>/dev/null && echo "Pro available" || echo "OSS only"
-```
+**Optional:** Semgrep Pro — enables cross-file taint tracking, inter-procedural analysis, and additional languages (Apex, C#, Elixir). Read local installation metadata if available and record Pro as installed, absent or unverified. Do not use a registry ruleset or a Pro/license/network probe as an availability check before approval. Pro usability is established only by authorized execution; a failed probe is not evidence of OSS-only availability.
 
 **Limitations:** OSS mode cannot track data flow across files. Pro mode uses `-j 1` for cross-file analysis (slower per ruleset, but parallel rulesets compensate).
+The bundled scan-status helper requires a supported Bash shell and local Python 3 to validate each fresh JSON/SARIF pair. Check these local capabilities before execution; missing tooling does not authorize installation.
 
 ## Scan Modes
 
@@ -109,10 +105,10 @@ See [scan-modes.md](references/scan-modes.md) for metadata criteria and jq filte
 ```
 ┌──────────────────────────────────────────────────────────────────┐
 │ MAIN AGENT (this skill)                                          │
-│ Step 1: Detect languages + check Pro availability                │
+│ Step 1: Detect languages + inspect local engine evidence        │
 │ Step 2: Select scan mode + rulesets (ref: rulesets.md)           │
 │ Step 3: Present plan + rulesets, get approval [⛔ HARD GATE]     │
-│ Step 4: Spawn parallel scan Tasks (approved rulesets + mode)     │
+│ Step 4: Execute approved scans via workers or directly          │
 │ Step 5: Merge results and report                                 │
 └──────────────────────────────────────────────────────────────────┘
          │ Step 4
@@ -134,41 +130,37 @@ See [scan-modes.md](references/scan-modes.md) for metadata criteria and jq filte
 
 | Step | Action | Gate | Key Reference |
 |------|--------|------|---------------|
-| 1 | Resolve output dir, detect languages + Pro availability | — | Use Glob, not Bash |
+| 1 | Resolve output dir, detect languages + local engine evidence | Local only | Supported file search and local metadata |
 | 2 | Select scan mode + rulesets | — | [rulesets.md](references/rulesets.md) |
-| 3 | Present plan, get explicit approval | ⛔ HARD | AskUserQuestion |
-| 4 | Spawn parallel scan Tasks | — | [scanner-task-prompt.md](references/scanner-task-prompt.md) |
+| 3 | Present plan, get explicit approval | ⛔ HARD | Supported user-input tool or ordinary chat |
+| 4 | Execute approved scans via supported workers or directly | Approved actions only | [scanner-task-prompt.md](references/scanner-task-prompt.md) |
 | 5 | Merge results and report | — | Merge script (below) |
 
-**Task enforcement:** On invocation, create 5 tasks with blockedBy dependencies (each step blocks the previous). Step 3 is a HARD GATE — mark complete ONLY after user explicitly approves.
+**Task tracking:** Use the host's supported task tracker for the five ordered steps when available, with each step depending on the previous. Otherwise keep a short checklist. Step 3 is a HARD GATE regardless of tracking tools; mark it complete only with explicit authorization for the exact plan.
 
 **Merge command (Step 5):**
 
 ```bash
 uv run {baseDir}/scripts/merge_sarif.py $OUTPUT_DIR/raw $OUTPUT_DIR/results/results.sarif
 ```
 
-## Agents
-
-| Agent | Tools | Purpose |
-|-------|-------|---------|
-| `static-analysis:semgrep-scanner` | Bash | Executes parallel semgrep scans for a language category |
+## Execution capability
 
-Use `subagent_type: static-analysis:semgrep-scanner` in Step 4 when spawning Task subagents.
+Use an available general-purpose worker with shell/file access and the bounded [scanner task prompt](references/scanner-task-prompt.md) when delegation is authorized. A specialist scanner is optional only when the current host actually exposes it; the repository bundles no exact scanner-agent dependency. If workers are unavailable or delegation is not authorized, use direct execution of the same approved task with bounded concurrency. Report the execution path actually used.
 
 ## Rationalizations to Reject
 
 | Shortcut | Why It's Wrong |
 |----------|----------------|
-| "User asked for scan, that's approval" | Original request ≠ plan approval. Present plan, use AskUserQuestion, await explicit "yes" |
+| "User asked for scan, that's approval" | A generic scan request does not specify a plan. Obtain missing exact-plan authority; preserve an existing explicit approval |
 | "Step 3 task is blocking, just mark complete" | Lying about task status defeats enforcement. Only mark complete after real approval |
 | "I already know what they want" | Assumptions cause scanning wrong directories/rulesets. Present plan for verification |
 | "Just use default rulesets" | User must see and approve exact rulesets before scan |
 | "Add extra rulesets without asking" | Modifying approved list without consent breaks trust |
 | "Third-party rulesets are optional" | Trail of Bits, 0xdea, Decurity catch vulnerabilities not in official registry — REQUIRED |
 | "Use --config auto" | Sends metrics; less control over rulesets |
-| "One Task at a time" | Defeats parallelism; spawn all Tasks together |
-| "Pro is too slow, skip --pro" | Cross-file analysis catches 250% more true positives; worth the time |
+| "A missing scanner agent blocks every scan" | Use a supported general worker when authorized, or execute the bounded approved tasks directly |
+| "Change the engine without asking" | Pro and OSS have different coverage; preserve the approved engine or obtain a changed plan |
 | "Semgrep handles GitHub URLs natively" | URL handling fails on repos with non-standard YAML; always clone first |
 | "Cleanup is optional" | Cloned repos pollute the user's workspace and accumulate across runs |
 | "Use `.` or relative path as target" | Subagents need absolute paths to avoid ambiguity |
@@ -190,11 +182,12 @@ Use `subagent_type: static-analysis:semgrep-scanner` in Step 4 when spawning Tas
 
 - [ ] Output directory resolved (user-specified or auto-incremented default)
 - [ ] All generated files stored inside `$OUTPUT_DIR`
-- [ ] Languages detected with file counts; Pro status checked
+- [ ] Languages detected with file counts; local engine evidence and Pro uncertainty recorded
 - [ ] Scan mode selected by user (run all / important only)
 - [ ] Rulesets include third-party rules for all detected languages
 - [ ] User explicitly approved the scan plan (Step 3 gate passed)
-- [ ] All scan Tasks spawned in a single message and completed
+- [ ] Approved scans completed via supported authorized workers or bounded direct execution
+- [ ] Every approved scan's PID/status and fresh JSON/SARIF pair passed `wait_semgrep_scans`; no incomplete batch or successful subset is reported as complete
 - [ ] Every `semgrep` command used `--metrics=off`
 - [ ] Approved rulesets logged to `$OUTPUT_DIR/rulesets.txt`
 - [ ] Raw per-scan outputs stored in `$OUTPUT_DIR/raw/`

@@ -138,16 +138,16 @@ These are **NOT optional**. Include automatically when language matches:
 | Java | [Atlassian Labs](https://github.com/atlassian-labs/atlassian-sast-ruleset) | Atlassian-maintained Java rules |
 | Python, JS/TS, Java, Ruby, Go, PHP | [Apiiro](https://github.com/apiiro/malicious-code-ruleset) | Malicious code detection, supply chain |
 
-### Step 5: Verify Rulesets
+### Step 5: Propose Ruleset Verification
 
-Before finalizing, verify official rulesets load:
+Before approval, record proposed ruleset identifiers without contacting registries. Registry validation and browsing are network actions: list the exact lookups in the Step 3 plan and perform them only after authorization. With that authority, verify approved rulesets load and retain the command's actual status:
 
 ```bash
-# Quick validation (exits 0 if valid)
-semgrep --config p/python --validate --metrics=off 2>&1 | head -3
+# After Step 3 authorizes this registry lookup (do not pipe away its status):
+SEMGREP_ENABLE_VERSION_CHECK=0 semgrep --config p/python --validate --metrics=off
 ```
 
-Or browse the [Semgrep Registry](https://semgrep.dev/explore).
+The [Semgrep Registry](https://semgrep.dev/explore) is an alternative after the same network authorization. For an approved local-only scan, validate only the approved local rule paths and report registry coverage as unverified.
 
 ### Output Format
 

@@ -102,9 +102,11 @@ done
 In important-only mode, add `[SEVERITY_FLAGS]` to the scanner template:
 
 ```bash
-semgrep [--pro if available] --metrics=off [SEVERITY_FLAGS] --config [RULESET] --json -o [OUTPUT_DIR]/raw/[lang]-[ruleset].json --sarif-output=[OUTPUT_DIR]/raw/[lang]-[ruleset].sarif [TARGET] &
+start_semgrep_scan [OUTPUT_DIR]/raw/[lang]-[ruleset].json [OUTPUT_DIR]/raw/[lang]-[ruleset].sarif env SEMGREP_ENABLE_VERSION_CHECK=0 semgrep [--pro if approved] --metrics=off [SEVERITY_FLAGS] --config [RULESET] --json -o [OUTPUT_DIR]/raw/[lang]-[ruleset].json --sarif-output=[OUTPUT_DIR]/raw/[lang]-[ruleset].sarif [TARGET] || SEMGREP_SCAN_LAUNCH_FAILED=1
 ```
 
+Source the scan-status helper and require `wait_semgrep_scans` to pass as described in [scanner-task-prompt.md](scanner-task-prompt.md). Severity filtering never waives process-status or expected-artifact checks.
+
 Where `[SEVERITY_FLAGS]` is:
 - **Run all**: *(empty)*
 - **Important only**: `--severity MEDIUM --severity HIGH --severity CRITICAL`
-Original file line number
+Diff line change
@@ Expand Up @@
     ## Gate
-    For UI/browser/styling/routing/reduced-motion/forced-colors changes, run `npm run ensure` before browser QA (never assume port 3000/3001/3002), then use `npm run verify:ui` (Chromium) as the gate. These are local/offline-safe (demo mode).
+    Follow `docs/agents/verification-gates.md#gate-selection`. For UI/browser/styling/routing/reduced-motion/forced-colors changes, run `npm run ensure` before browser QA, use its printed URL and verify `/api/local-project-id`, then prove the changed owner or journey first. Use `npm run verify:ui` for shared UI foundations or when PR/handoff policy requires the complete Chromium gate. Report focused proof as focused proof, retaining required broad gates and any physical Safari or installed-PWA acceptance gaps.
     ## Review Checklist
@@ Expand Down @@