workflow-code-to-spec
Pass
Audited by Gen Agent Trust Hub on Nov 10, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to execute a local utility script using
node .claude/scripts/lib/workflow-skill-contract.cjs --write. This is used to regenerate workflow documentation blocks based on registry edits, which is a standard automation pattern within the project environment. - [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process external data including implementation code, registry files (
.claude/workflows.json), and project configurations (docs/project-config.json). This creates a surface for indirect prompt injection where malicious instructions hidden in the analyzed codebase could attempt to influence the agent's output or behavior. - Ingestion points: Implementation source code,
.claude/workflows.json, anddocs/project-config.json. - Boundary markers: The skill uses a 'Coverage Ledger' and 'Goal Contract' to track and validate work, though specific prompt-level delimiters to neutralize embedded instructions in source code are not explicitly defined in this file.
- Capability inventory: Capability to write files (specifications and reports), execute local scripts via
node, and spawn sub-agents usingspawn_agent. - Sanitization: No explicit sanitization or filtering of the ingested source code content is described before processing.
Audit Metadata