Menu
Tracker-specific security checklist. Use when modifying actor classification, approval provenance, description integrity, blocked paths, dispatch logic, or MCP tool authorization.
Installer avec Codex ou Claude Copiez ce prompt, collez-le dans Codex, Claude ou un autre assistant, puis laissez-le vérifier la page du skill et l'installer pour vous.
Basé sur la classification professionnelle SOC
Use for any question about a codebase, its architecture, file relationships, or project content — especially when graphify-out/ exists, where the question should be treated as a graphify query first. Turns any input (code, docs, papers, images, videos) into a persistent knowledge graph with god nodes, community detection, and query/path/explain tools.
Guide for developing new MCP tools. Use when adding tools to mcp-server.ts or via space plugin mcpTools arrays. Covers Zod validation, actor handling, and test patterns.
Database patterns for better-sqlite3 with WAL mode. Query optimization, schema design, and migration patterns for the tracker's SQLite layer.
Safety guidelines for modifying orchestrator code. Use when touching state transitions, SSE handling, session lifecycle, circuit breaker, safe restart, or dispatch logic in orchestrator.ts.
Research-before-coding workflow. Search for existing patterns, helpers, and implementations in the tracker codebase before writing new code.
Use when adding authentication, handling user input, working with secrets, creating API endpoints, or modifying security-sensitive code. General security checklist.
| name | tracker-security-review |
| description | Tracker-specific security checklist. Use when modifying actor classification, approval provenance, description integrity, blocked paths, dispatch logic, or MCP tool authorization. |
| origin | custom |
classifyActor() or actor classification logicgetDispatchableItems() or dispatch eligibilityconfig.tsclassifyActor() in db.ts correctly classifies all actor nameshuman-class actors can move items to approved or cancelledcreated_by to the agent name (prevent impersonation)actor_class = "agent"requires_code=0) may be agent-approvedapproved_by records the actual actor nameapproved_by_class is set and verified at dispatchapproved_at timestamp is recordedapproved_description_hash is SHA-256 of description at approval timegetDispatchableItems() verifies approved_by_class = 'human' (except comment-only)BLOCKED_PATHS in config.ts includes all security-critical filesbuildPrompt())buildPrompt() includes security rules for coder bots