claude-md-updater
Pass
Audited by Gen Agent Trust Hub on Oct 31, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes conversation history to generate persistent project instructions, creating a surface for indirect injection.
- Ingestion points: The entire conversation history is scanned for keywords and durable facts (SKILL.md).
- Boundary markers: The instructions do not define specific delimiters or instructions to ignore embedded commands within the conversation history during the scanning process.
- Capability inventory: The skill proposes unified diffs for
AGENTS.mdandCLAUDE.md, which are loaded into future agent sessions to influence behavior (SKILL.md). - Sanitization: No automated sanitization is described, but the skill explicitly requires the user to approve the exact diff before writing any changes to the filesystem.
Audit Metadata