agents-md-convert
Pass
Audited by Gen Agent Trust Hub on Sep 24, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from repository instruction files and their functional consumers to refactor guidance ownership.
- Ingestion points: The skill reads
AGENTS.md,CLAUDE.md, and local override files (AGENTS.override.md,CLAUDE.local.md) from a Git repository, as well as functional consumers like build scripts or configuration files that handle instructions (Section 2). - Boundary markers: There are no explicit instructions for the agent to use specific delimiters or to disregard instructions embedded within the processed data during the conversion process.
- Capability inventory: The skill allows for file system modifications (
Applymode in Section 4), Git commands, and the execution of external CLI tools ("harness probes" in Section 6). - Sanitization: The instructions do not define sanitization or validation logic for the content of the files being moved or modified.
- [COMMAND_EXECUTION]: The skill involves the execution of local shell commands for Git operations and environment auditing.
- Evidence: Section 6 ("Offer live canaries") describes running installed harnesses with
--helpand selecting sentinels to probe discovery behavior. Sections 1 and 2 involve Git operations to resolve boundaries and enumerate files.
Audit Metadata