consult-decisions
Pass
Audited by Gen Agent Trust Hub on Sep 29, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to treat markdown files in
docs/adr/anddocs/pdr/as authoritative constraints for planning, designing, and implementing work. This creates a surface where an attacker who can modify project documentation could influence the agent's behavior via embedded instructions. - Ingestion points: Project documentation directories
docs/adr/anddocs/pdr/(SKILL.md). - Boundary markers: Absent; the skill does not instruct the agent to ignore or delimit embedded instructions within the records.
- Capability inventory: File reading and influence over critical agent tasks including design, planning, implementation, and code changes.
- Sanitization: Absent; the content of the records is processed as authoritative without validation or filtering.
- [COMMAND_EXECUTION]: The skill provides a shell script using
find,sort, andawkto extract metadata from decision record frontmatter. While the command itself is a standard utility for cataloging, it demonstrates the ability to execute shell commands to process file content.
Audit Metadata