@matyah00/openpi 0.1.4 → 0.2.0

package/README.md

@@ -84,6 +84,8 @@ The npm names `openpi` and `open-pi` are blocked by npm's package-name similarit
 /openpi use commands         enable prompt commands and core tools
 /openpi use workflow         enable /add, /fix, /review and spawn_agents
 /openpi use guard            enable security, dependency, and ship gates
+/openpi use commands+guard   combine multiple profiles (e.g. focus+safety)
+/openpi use full --dry-run   preview profile settings changes without writing
 /openpi use full             enable the broad OpenPi surface
 /openpi clear                remove OpenPi-managed extensions
 ```
@@ -102,12 +104,13 @@ Profiles update `.pi/settings.json`. Run `/reload` or restart Pi after switching
 
 ```text
 OpenPi
-  13 profiles          choose a focused runtime surface per task
-  19 prompt commands   /prime, /blueprint, /deep, /ship, /parallel, ...
-  36 agent prompts     planner, reviewer, tester, security-auditor, Pi experts
-  7 skills             ultrathink, test-first, security-guard, bowser, ...
+  13 profiles          choose or compose a focused runtime surface
+  23 prompt commands   /prime, /blueprint, /deep, /ship, /refactor, ...
+  42 agent prompts     planner, reviewer, tester, api-designer, Pi experts
+  9 skills             ultrathink, test-first, refactor-guide, bowser, ...
   11 themes            tokyo-night, rose-pine, gruvbox, nord, dracula, ...
   Native tools         search, audit, state, snapshot, dispatch, chains
+  Audit Telemetry      structured JSONL audit logging to .pi/logs/
 ```
 
 OpenPi is designed around one practical rule: load only the surface you need.
@@ -195,6 +198,10 @@ OpenPi role agents include:
 | `spec-reviewer` | Challenges unclear requirements before implementation |
 | `ship-guard` | Reviews release readiness |
 | `red-team` | Challenges plans and assumptions |
+| `api-designer` | Reviews REST/GraphQL APIs for design and naming consistency |
+| `perf-auditor` | Evaluates bundle sizes, N+1 queries, and resource leaks |
+| `migration-expert` | Plans database and architecture migration sequences |
+| `docs-writer` | Generates/maintains READMEs, manuals, and architecture documents |
 
 The `spawn_agents` tool can run agents sequentially or in parallel as isolated Pi subprocesses. It returns structured outputs: files, line ranges, commands, exact validation output, findings, and assumptions.
 
@@ -222,6 +229,10 @@ The `team` profile registers `dispatch_agent` and uses teams from `agents/teams.
 | `frontend` | scout, frontend, reviewer |
 | `backend` | scout, backend, reviewer |
 | `pi-pi` | Pi package, extension, skill, prompt, config, theme, TUI, CLI, and keybinding experts |
+| `perf` | scout, perf-auditor, planner, reviewer |
+| `docs` | scout, docs-writer, reviewer |
+| `api` | scout, api-designer, reviewer |
+| `migration` | scout, migration-expert, planner, reviewer |
 
 ### Chain runner
 
@@ -241,6 +252,10 @@ The `chain` profile registers `run_chain` for sequential workflows from `agents/
 | `spec-to-plan` | problem-architect -> spec-reviewer -> planner |
 | `ship-gate` | security-auditor -> ship-guard -> reviewer |
 | `pi-package-design` | Pi experts -> planner |
+| `refactor-flow` | scout -> problem-architect -> planner -> reviewer |
+| `docs-flow` | scout -> docs-writer -> reviewer |
+| `perf-audit` | scout -> perf-auditor -> planner |
+| `full-lifecycle` | problem-architect -> planner -> plan-reviewer -> builder -> tester -> reviewer |
 
 ---
 
@@ -250,19 +265,20 @@ OpenPi registers native Pi tools through profiles.
 
 | Tool | Purpose |
 |------|---------|
-| `project_tree` | Return a scoped project tree with ignore handling |
-| `code_search_batch` | Run multiple code searches in one call |
-| `env_scan` | Detect stack, package managers, scripts, and environment clues |
-| `secret_scan` | Search for common secret and credential patterns |
-| `ghost_test_scan` | Find weak, vacuous, or reward-hacked tests |
+| `project_tree` | Return a scoped project tree with ignore handling (supports file sizes) |
+| `code_search_batch` | Run multiple code searches in one call (supports case-sensitivity & regex toggle) |
+| `env_scan` | Detect stack, package managers, scripts, and environment clues (24+ frameworks) |
+| `secret_scan` | Search for 22 types of cloud keys, tokens, credentials, and high-entropy hex strings |
+| `ghost_test_scan` | Find weak, empty, mocked-everything, self-comparing, or reward-hacked tests |
+| `sast_scan` | Static Application Security Testing for eval, SQL concat, SSRF, CORS wildcards, and XSS |
 | `dependency_inventory` | Summarize dependency manifests and lockfiles |
 | `session_state` | Read current session state |
 | `goal_state` | Read goal memory state |
 | `write_snapshot` | Write a continuation snapshot |
 | `parallel_safety_check` | Check file ownership overlap before parallel work |
-| `spawn_agents` | Run role agents as isolated Pi subprocesses |
+| `spawn_agents` | Run role agents as isolated Pi subprocesses (supports retries & timeouts) |
 | `dispatch_agent` | Dispatch to the active specialist team |
-| `run_chain` | Run a named sequential agent chain |
+| `run_chain` | Run a named sequential agent chain (supports step timeouts, continueOnError, & $STEP_N) |
 
 ---
 
@@ -279,6 +295,8 @@ OpenPi ships focused Pi skills:
 | `session-continuity` | Long context, resuming, stopping, compaction handoffs |
 | `env-scanner` | Unknown repos, setup issues, stack detection |
 | `bowser` | Playwright-powered browser automation and UI testing |
+| `refactor-guide` | Guides safe codebase refactoring with code smell detection and verification |
+| `perf-auditor` | Structured resource profiling and runtime execution optimization |
 
 ---
 

package/agents/agent-chain.yaml

@@ -111,3 +111,53 @@ backend-flow:
       prompt: "Implement or design the backend change using this context:\n\n$INPUT"
     - agent: reviewer
       prompt: "Review the backend result for API, persistence, auth, and test risks:\n\n$INPUT"
+
+refactor-flow:
+  description: "Explore, analyze, plan refactoring, then review safety"
+  steps:
+    - agent: scout
+      prompt: "Map the code areas relevant to this refactoring: $INPUT"
+    - agent: problem-architect
+      prompt: "Analyze the code smell and structural issues. Identify the minimal safe refactoring:\n\n$INPUT"
+    - agent: planner
+      prompt: "Create a step-by-step refactoring plan with verification at each step:\n\n$INPUT\n\nOriginal request: $ORIGINAL"
+    - agent: reviewer
+      prompt: "Review whether this refactoring plan preserves behavior and has adequate verification:\n\n$INPUT"
+
+docs-flow:
+  description: "Explore project, generate documentation, then review accuracy"
+  steps:
+    - agent: scout
+      prompt: "Map the project structure, public APIs, and existing documentation for: $INPUT"
+    - agent: docs-writer
+      prompt: "Generate or update the documentation using this project analysis:\n\n$INPUT\n\nOriginal request: $ORIGINAL"
+    - agent: reviewer
+      prompt: "Review the documentation for accuracy, completeness, and clarity:\n\n$INPUT"
+
+perf-audit:
+  description: "Explore, audit performance, then plan improvements"
+  steps:
+    - agent: scout
+      prompt: "Map the code areas and hot paths relevant to: $INPUT"
+    - agent: perf-auditor
+      prompt: "Run a performance audit on the mapped code areas:\n\n$INPUT\n\nOriginal request: $ORIGINAL"
+    - agent: planner
+      prompt: "Create a prioritized improvement plan from the performance audit:\n\n$INPUT\n\nOriginal request: $ORIGINAL"
+
+full-lifecycle:
+  description: "Complete feature lifecycle: scope, plan, critique, revise, build, test, review"
+  steps:
+    - agent: problem-architect
+      prompt: "Analyze scope, risks, and required context for: $INPUT"
+    - agent: planner
+      prompt: "Create a detailed implementation plan:\n\n$INPUT\n\nOriginal request: $ORIGINAL"
+    - agent: plan-reviewer
+      prompt: "Critically review this plan for missing steps, risks, and ordering:\n\n$INPUT"
+    - agent: planner
+      prompt: "Revise the plan using this critique:\n\n$INPUT\n\nOriginal request: $ORIGINAL"
+    - agent: builder
+      prompt: "Implement the revised plan:\n\n$INPUT"
+    - agent: tester
+      prompt: "Run targeted validation for the implementation:\n\n$INPUT"
+    - agent: reviewer
+      prompt: "Final review of the implementation for bugs, regressions, and missing tests:\n\n$INPUT\n\nOriginal request: $ORIGINAL"

package/agents/api-designer.md

@@ -0,0 +1,58 @@
+---
+name: api-designer
+description: Reviews and designs REST/GraphQL APIs for consistency, naming conventions, versioning, error handling, pagination, and idempotency.
+tools: read, grep, find, ls, code_search_batch
+thinking: medium
+---
+
+You are an API design reviewer for a Pi coding workflow. You never edit files.
+
+## Review Areas
+
+1. **Naming consistency**: resource naming, plural vs singular, casing conventions.
+2. **HTTP method correctness**: GET for reads, POST for creates, PUT/PATCH for updates, DELETE for removes.
+3. **Error handling**: consistent error response shapes, appropriate status codes, actionable error messages.
+4. **Versioning**: URL path vs header versioning, backward compatibility.
+5. **Pagination**: cursor vs offset, consistent response envelope.
+6. **Idempotency**: POST/PUT idempotency keys, retry safety.
+7. **Authentication/Authorization**: consistent auth patterns, proper 401 vs 403 usage.
+8. **Rate limiting**: headers, retry-after, graceful degradation.
+9. **Input validation**: request body validation, query parameter constraints.
+10. **Documentation**: OpenAPI/Swagger completeness, example payloads.
+
+## Process
+
+- Use `code_search_batch` to find route definitions, controller handlers, and middleware.
+- Use `read` to inspect request/response shapes and error handling.
+- Use `grep` to find inconsistencies in naming or status codes.
+
+## Output
+
+Return:
+
+```text
+## API Review: {scope}
+
+### Endpoints Reviewed
+- METHOD /path - summary
+
+### Consistency Score: N/10
+
+### Issues (by severity)
+1. [CRITICAL] ...
+2. [HIGH] ...
+3. [MEDIUM] ...
+4. [LOW] ...
+
+### Naming Conventions
+- Current: ...
+- Recommendation: ...
+
+### Error Handling
+- Current pattern: ...
+- Missing: ...
+
+### Recommendations
+1. ...
+2. ...
+```

package/agents/docs-writer.md

@@ -0,0 +1,38 @@
+---
+name: docs-writer
+description: Generates and updates documentation — README, API docs, architecture guides, changelogs, and inline comments — from code analysis.
+tools: read, grep, find, ls, code_search_batch, project_tree
+thinking: medium
+---
+
+You are a documentation writer for a Pi coding workflow.
+
+## Documentation Types
+
+1. **README**: Project overview, setup instructions, usage examples, contribution guide.
+2. **API documentation**: Endpoint descriptions, request/response examples, auth requirements.
+3. **Architecture docs**: System overview, component relationships, data flow, decision records.
+4. **Changelog entries**: What changed, why, migration steps if breaking.
+5. **Inline comments**: Complex logic explanations, type annotations, usage examples.
+
+## Process
+
+1. Use `project_tree` to understand project structure.
+2. Use `code_search_batch` to find exports, public APIs, route definitions.
+3. Use `read` to understand implementation details.
+4. Match documentation style to existing docs if present.
+5. Use concrete code examples, not abstract descriptions.
+
+## Rules
+
+- Write for the next developer, not the current one.
+- Include working code examples that can be copy-pasted.
+- Document error cases and edge cases, not just happy paths.
+- Keep language precise and concise.
+- Use consistent heading hierarchy.
+- Include prerequisites and setup steps.
+- Add links between related documentation sections.
+
+## Output
+
+Return the documentation in the appropriate format (markdown for docs, JSDoc/docstrings for inline).

package/agents/migration-expert.md

@@ -0,0 +1,62 @@
+---
+name: migration-expert
+description: Plans database, API, framework, and dependency migrations with rollback strategies, risk assessment, and phased execution.
+tools: read, grep, find, ls, code_search_batch, env_scan, dependency_inventory
+thinking: high
+---
+
+You are a migration planning expert for a Pi coding workflow. You never edit files unless the parent explicitly asks for implementation.
+
+## Migration Types
+
+1. **Database migrations**: Schema changes, data transformations, index additions/removals.
+2. **Framework migrations**: Major version upgrades, framework switches (e.g. Express→Fastify).
+3. **API migrations**: Breaking changes, versioning transitions, deprecation paths.
+4. **Dependency migrations**: Major version bumps, package replacements, security patches.
+5. **Infrastructure migrations**: Cloud provider changes, containerization, serverless transitions.
+
+## Analysis Process
+
+1. Identify the current state with `env_scan` and `dependency_inventory`.
+2. Map all touchpoints with `code_search_batch`.
+3. Identify breaking changes and backward compatibility requirements.
+4. Design rollback strategy for each phase.
+5. Estimate risk and effort per phase.
+
+## Output
+
+Return:
+
+```text
+## Migration Plan: {from} → {to}
+
+### Current State
+- Version/framework: ...
+- Dependent code: N files, N modules
+- Test coverage: ...
+
+### Breaking Changes
+1. ...
+
+### Phases
+#### Phase 1: Preparation (rollback: trivial)
+- [ ] ...
+
+#### Phase 2: Migration (rollback: revert commit)
+- [ ] ...
+
+#### Phase 3: Cleanup (rollback: N/A)
+- [ ] ...
+
+### Rollback Strategy
+- Before point of no return: ...
+- After point of no return: ...
+
+### Risk Assessment
+- Overall: LOW | MEDIUM | HIGH
+- Data loss risk: ...
+- Downtime: ...
+
+### Validation
+- ...
+```

package/agents/perf-auditor.md

@@ -0,0 +1,64 @@
+---
+name: perf-auditor
+description: Read-only performance auditor that identifies bundle bloat, N+1 queries, missing indexes, memory leaks, unnecessary re-renders, and slow patterns.
+tools: read, grep, find, ls, code_search_batch, bash
+thinking: high
+---
+
+You are a read-only performance auditor for a Pi coding workflow. You never edit files.
+
+## Audit Areas
+
+### Frontend
+- Bundle size: large imports, missing tree-shaking, barrel file re-exports.
+- Re-render storms: missing memoization, inline object/function props, unstable keys.
+- Layout thrashing: synchronous DOM reads/writes, forced reflows.
+- Asset loading: unoptimized images, missing lazy loading, blocking scripts.
+
+### Backend
+- N+1 queries: ORM loops that generate one query per iteration.
+- Missing database indexes: queries filtering on unindexed columns.
+- Memory leaks: unclosed connections, growing caches, event listener accumulation.
+- Blocking operations: synchronous file I/O, CPU-bound work on event loop.
+- Connection pooling: missing or misconfigured pools.
+
+### General
+- Algorithmic complexity: O(n²) or worse in hot paths.
+- Unnecessary serialization: repeated JSON parse/stringify cycles.
+- Missing caching: repeated identical computations or network calls.
+- Cold start: heavy initialization, unnecessary eager loading.
+
+## Process
+
+- Use `code_search_batch` for ORM query patterns, import sizes, memo usage.
+- Use `bash` for read-only commands: `du -sh`, `wc -l`, dependency size checks.
+- Use `grep` for known anti-patterns.
+
+## Output
+
+Return:
+
+```text
+## Performance Audit: {scope}
+
+### Risk Level: LOW | MEDIUM | HIGH | CRITICAL
+
+### Findings (by impact)
+1. [CRITICAL] file:line - description - estimated impact
+2. [HIGH] ...
+3. [MEDIUM] ...
+
+### Bundle Analysis
+- Entry points: ...
+- Largest dependencies: ...
+- Tree-shaking opportunities: ...
+
+### Query Analysis
+- N+1 patterns found: N
+- Missing indexes: ...
+
+### Recommendations
+1. Quick wins (< 1 hour): ...
+2. Medium effort (1-4 hours): ...
+3. Architecture changes: ...
+```

package/agents/teams.yaml

@@ -4,6 +4,7 @@ full:
   - builder
   - reviewer
   - documenter
+  - docs-writer
   - red-team
 
 plan-build:
@@ -60,6 +61,28 @@ backend:
   - backend
   - reviewer
 
+perf:
+  - scout
+  - perf-auditor
+  - planner
+  - reviewer
+
+docs:
+  - scout
+  - docs-writer
+  - reviewer
+
+api:
+  - scout
+  - api-designer
+  - reviewer
+
+migration:
+  - scout
+  - migration-expert
+  - planner
+  - reviewer
+
 pi-pi:
   - pi-orchestrator
   - ext-expert

package/damage-control-rules.yaml

@@ -17,6 +17,7 @@ bashToolPatterns:
     reason: recursive chmod 777
   - pattern: '\bchown\s+-[Rr].*\broot\b'
     reason: recursive chown to root
+  # --- Git destructive operations ---
   - pattern: '\bgit\s+reset\s+--hard\b'
     reason: git reset --hard (use --soft or stash)
   - pattern: '\bgit\s+clean\s+(-[^\s]*)*-[fd]'
@@ -51,6 +52,7 @@ bashToolPatterns:
   - pattern: '\bgit\s+push\s+\S+\s+:\S+'
     reason: Deletes remote branch (old syntax)
     ask: true
+  # --- System-level destructive ---
   - pattern: '\bmkfs\.'
     reason: filesystem format command
   - pattern: '\bdd\s+.*of=/dev/'
@@ -63,6 +65,7 @@ bashToolPatterns:
     reason: pkill -9
   - pattern: '\bhistory\s+-c\b'
     reason: clearing shell history
+  # --- AWS ---
   - pattern: '\baws\s+s3\s+rm\s+.*--recursive'
     reason: aws s3 rm --recursive (deletes all objects)
   - pattern: '\baws\s+s3\s+rb\s+.*--force'
@@ -83,6 +86,7 @@ bashToolPatterns:
     reason: aws iam delete-role
   - pattern: '\baws\s+iam\s+delete-user\b'
     reason: aws iam delete-user
+  # --- GCP ---
   - pattern: '\bgcloud\s+projects\s+delete\b'
     reason: gcloud projects delete (DELETES ENTIRE PROJECT)
   - pattern: '\bgcloud\s+compute\s+instances\s+delete\b'
@@ -108,6 +112,7 @@ bashToolPatterns:
   - pattern: '\bgcloud\s+iam\s+policies\b'
     reason: gcloud iam policies (modifies IAM policies)
     ask: true
+  # --- Firebase ---
   - pattern: '\bfirebase\s+projects:delete\b'
     reason: firebase projects:delete (deletes entire project)
   - pattern: '\bfirebase\s+firestore:delete\s+.*--all-collections'
@@ -118,6 +123,7 @@ bashToolPatterns:
     reason: firebase hosting:disable
   - pattern: '\bfirebase\s+functions:delete\b'
     reason: firebase functions:delete
+  # --- Vercel ---
   - pattern: '\bvercel\s+remove\s+.*--yes'
     reason: vercel remove --yes (removes deployment)
   - pattern: '\bvercel\s+projects\s+rm\b'
@@ -130,10 +136,12 @@ bashToolPatterns:
     reason: vercel remove (removes deployment)
   - pattern: '\bvercel\s+domains\s+rm\b'
     reason: vercel domains rm (removes custom domain)
+  # --- Netlify ---
   - pattern: '\bnetlify\s+sites:delete\b'
     reason: netlify sites:delete (deletes entire site)
   - pattern: '\bnetlify\s+functions:delete\b'
     reason: netlify functions:delete
+  # --- Cloudflare ---
   - pattern: '\bwrangler\s+delete\b'
     reason: wrangler delete (deletes Worker)
   - pattern: '\bwrangler\s+r2\s+bucket\s+delete\b'
@@ -144,6 +152,112 @@ bashToolPatterns:
     reason: wrangler d1 delete (deletes database)
   - pattern: '\bwrangler\s+queues\s+delete\b'
     reason: wrangler queues delete
+  # --- DigitalOcean ---
+  - pattern: '\bdoctl\s+compute\s+droplet\s+delete\b'
+    reason: doctl compute droplet delete (destroys server)
+  - pattern: '\bdoctl\s+databases\s+delete\b'
+    reason: doctl databases delete (destroys database cluster)
+  - pattern: '\bdoctl\s+kubernetes\s+cluster\s+delete\b'
+    reason: doctl kubernetes cluster delete (destroys k8s cluster)
+  - pattern: '\bdoctl\s+apps\s+delete\b'
+    reason: doctl apps delete (destroys App Platform app)
+  - pattern: '\bdoctl\s+compute\s+volume\s+delete\b'
+    reason: doctl compute volume delete (destroys block storage)
+  # --- Heroku ---
+  - pattern: '\bheroku\s+apps:destroy\b'
+    reason: heroku apps:destroy (permanently deletes app)
+  - pattern: '\bheroku\s+pg:reset\b'
+    reason: heroku pg:reset (wipes entire database)
+  - pattern: '\bheroku\s+config:unset\b'
+    reason: heroku config:unset (removes environment variables)
+    ask: true
+  - pattern: '\bheroku\s+domains:remove\b'
+    reason: heroku domains:remove (removes custom domain)
+    ask: true
+  - pattern: '\bheroku\s+addons:destroy\b'
+    reason: heroku addons:destroy (removes addon and its data)
+  # --- Fly.io ---
+  - pattern: '\bfly\s+apps\s+destroy\b'
+    reason: fly apps destroy (permanently deletes app)
+  - pattern: '\bfly\s+volumes\s+destroy\b'
+    reason: fly volumes destroy (destroys persistent volume)
+  - pattern: '\bfly\s+secrets\s+unset\b'
+    reason: fly secrets unset (removes secrets)
+    ask: true
+  - pattern: '\bfly\s+postgres\s+destroy\b'
+    reason: fly postgres destroy (destroys database)
+  - pattern: '\bflyctl\s+destroy\b'
+    reason: flyctl destroy (permanently deletes app)
+  # --- Supabase ---
+  - pattern: '\bsupabase\s+db\s+reset\b'
+    reason: supabase db reset (wipes database and re-runs migrations)
+  - pattern: '\bsupabase\s+projects\s+delete\b'
+    reason: supabase projects delete (destroys project)
+  - pattern: '\bsupabase\s+functions\s+delete\b'
+    reason: supabase functions delete (removes edge function)
+  # --- Prisma ---
+  - pattern: '\bprisma\s+migrate\s+reset\b'
+    reason: prisma migrate reset (drops database and re-runs all migrations)
+  - pattern: '\bprisma\s+db\s+push\s+.*--force-reset'
+    reason: prisma db push --force-reset (drops and recreates database)
+  - pattern: '\bprisma\s+db\s+execute\b'
+    reason: prisma db execute (runs raw SQL against database)
+    ask: true
+  # --- Terraform ---
+  - pattern: '\bterraform\s+destroy\b'
+    reason: terraform destroy (destroys ALL managed infrastructure)
+  - pattern: '\bterraform\s+state\s+rm\b'
+    reason: terraform state rm (removes resource from state, may orphan infra)
+  - pattern: '\bterraform\s+taint\b'
+    reason: terraform taint (marks resource for recreation)
+    ask: true
+  - pattern: '\bterraform\s+force-unlock\b'
+    reason: terraform force-unlock (dangerous state lock override)
+  # --- Kubernetes ---
+  - pattern: '\bkubectl\s+delete\s+namespace\b'
+    reason: kubectl delete namespace (destroys ALL resources in namespace)
+  - pattern: '\bkubectl\s+delete\s+deployment\b'
+    reason: kubectl delete deployment
+    ask: true
+  - pattern: '\bkubectl\s+delete\s+pvc\b'
+    reason: kubectl delete pvc (destroys persistent volume claim and data)
+  - pattern: '\bkubectl\s+delete\s+secret\b'
+    reason: kubectl delete secret
+    ask: true
+  - pattern: '\bkubectl\s+delete\s+(-[^\s]*\s+)*--all\b'
+    reason: kubectl delete --all (mass deletion)
+  - pattern: '\bhelm\s+uninstall\b'
+    reason: helm uninstall (removes Helm release and its resources)
+    ask: true
+  - pattern: '\bhelm\s+delete\b'
+    reason: helm delete (removes Helm release)
+    ask: true
+  # --- Docker ---
+  - pattern: '\bdocker\s+system\s+prune\s+.*-a'
+    reason: docker system prune -a (removes ALL unused images, containers, networks)
+  - pattern: '\bdocker\s+volume\s+prune\b'
+    reason: docker volume prune (removes ALL unused volumes and their data)
+  - pattern: '\bdocker\s+rmi\s+.*-f\b'
+    reason: docker rmi -f (force removes images)
+  - pattern: '\bdocker\s+rm\s+.*-f\b'
+    reason: docker rm -f (force removes running containers)
+  - pattern: '\bdocker\s+compose\s+down\s+.*-v\b'
+    reason: docker compose down -v (removes volumes with data)
+    ask: true
+  # --- Pipe-to-shell attacks ---
+  - pattern: '\bcurl\s+.*\|\s*sh\b'
+    reason: Pipe-to-shell attack vector (curl | sh)
+  - pattern: '\bcurl\s+.*\|\s*bash\b'
+    reason: Pipe-to-shell attack vector (curl | bash)
+  - pattern: '\bwget\s+.*-O\s*-\s*\|\s*sh\b'
+    reason: Pipe-to-shell attack vector (wget -O- | sh)
+  - pattern: '\bwget\s+.*-O\s*-\s*\|\s*bash\b'
+    reason: Pipe-to-shell attack vector (wget -O- | bash)
+  - pattern: '\bcurl\s+.*\|\s*sudo\s+sh\b'
+    reason: Pipe-to-shell with sudo (extremely dangerous)
+  - pattern: '\bcurl\s+.*\|\s*sudo\s+bash\b'
+    reason: Pipe-to-shell with sudo (extremely dangerous)
+  # --- SQL ---
   - pattern: 'DELETE\s+FROM\s+\w+\s*;'
     reason: DELETE without WHERE clause (will delete ALL rows)
   - pattern: 'DELETE\s+\*\s+FROM'
@@ -159,6 +273,9 @@ bashToolPatterns:
   - pattern: '\bDELETE\s+FROM\s+\w+\s+WHERE\b.*\bid\s*='
     reason: SQL DELETE with specific ID
     ask: true
+  - pattern: '\bALTER\s+TABLE\s+\w+\s+DROP\s+COLUMN\b'
+    reason: ALTER TABLE DROP COLUMN (irreversible data loss)
+    ask: true
 
 zeroAccessPaths:
   - ".env"
@@ -202,6 +319,20 @@ zeroAccessPaths:
   - "dump.sql"
   - "backup.sql"
   - "*.dump"
+  # New zero-access paths
+  - "supabase/.env"
+  - ".wrangler/"
+  - "*.tfvars"
+  - "terraform.tfvars"
+  - ".pulumi/"
+  - "~/.config/heroku/"
+  - "~/.fly/"
+  - "~/.config/flyctl/"
+  - "~/.doctl/"
+  - "~/.config/doctl/"
+  - "~/.config/supabase/"
+  - "*.keystore"
+  - "*.jks"
 
 readOnlyPaths:
   - /etc/
@@ -245,6 +376,12 @@ readOnlyPaths:
   - .venv/
   - venv/
   - target/
+  # New read-only paths
+  - schema.prisma
+  - "drizzle/"
+  - "migrations/"
+  - "*.generated.ts"
+  - "*.generated.js"
 
 noDeletePaths:
   - CLAUDE.md
@@ -275,3 +412,19 @@ noDeletePaths:
   - docker-compose.yml
   - "docker-compose.*.yml"
   - .dockerignore
+  # New no-delete paths
+  - ".env.example"
+  - ".env.sample"
+  - "docker-compose.override.yml"
+  - Makefile
+  - Procfile
+  - "fly.toml"
+  - "vercel.json"
+  - "netlify.toml"
+  - "wrangler.toml"
+  - "supabase/config.toml"
+  - ".editorconfig"
+  - ".prettierrc"
+  - ".eslintrc"
+  - ".eslintrc.*"
+  - "biome.json"