agents-md-manager
Pass
Audited by Gen Agent Trust Hub on Oct 26, 2026
Risk Level: SAFECOMMAND_EXECUTIONDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to execute multiple Python scripts included in the package (analyze_project.py, convert_claude_to_agents.py, generate_config_toml.py, scaffold_codex_skill.py, validate_agents_md.py) to automate project management tasks.\n- [DATA_EXFILTRATION]: The conversion utility convert_claude_to_agents.py allows reading and inlining files from local paths, including support for home directory expansion via the ~ symbol in @import statements. This mechanism can be used to incorporate sensitive local file contents into the generated project documentation.\n- [INDIRECT_PROMPT_INJECTION]: The skill acts as a processing pipeline for project-specific metadata and documentation, which constitutes an attack surface for indirect prompt injection.\n
- Ingestion points: Multiple project configuration files (e.g., package.json, Cargo.toml, Makefile) and documentation files (CLAUDE.md, AGENTS.md) are read and processed.\n
- Boundary markers: Inlined content from external imports is merged into the output files without clear delimiters or safety instructions to prevent the agent from following instructions embedded in the data.\n
- Capability inventory: The skill scripts can perform directory scaffolding and file writing operations to create or update configuration and instruction files.\n
- Sanitization: There is no evidence of filtering or escaping external content before it is interpolated into the generated markdown or configuration files.
Audit Metadata