create-control-manifest
Pass
Audited by Gen Agent Trust Hub on Oct 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from various local repository files which could potentially contain malicious instructions if the repository itself is compromised.
- Ingestion points: The skill reads ADRs (
docs/architecture/adr-*.md), technical preferences (.claude/docs/technical-preferences.md), and engine reference documents (docs/engine-reference/) as specified inreferences/claude-skill.md. - Boundary markers: There are no explicit boundary markers or instructions to ignore embedded prompts within the files being read.
- Capability inventory: The skill has the ability to write files (
docs/architecture/control-manifest.md) and spawn sub-agents using theTasktool for technical reviews. - Sanitization: No automated sanitization or escaping of the content extracted from the source documents is performed before it is processed or written to the final manifest.
- [SAFE]: The skill follows established best practices for documentation generation, uses only platform-provided tools, and requires user approval before performing any write operations. All file access is scoped to the local repository.
Audit Metadata