plan-fix
Pass
Audited by Gen Agent Trust Hub on Sep 7, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted content from GitHub issues (title, body, comments) using the
gh issue viewcommand as described inreferences/PLANNING_PROCESS.mdto draft technical specifications. - Ingestion points:
references/PLANNING_PROCESS.mdfetches data from external GitHub issues including the title, body, and comments. - Boundary markers: The skill does not define explicit delimiters or instructions to treat issue content as untrusted data, which increases the risk that embedded instructions within an issue could influence the agent's logic during the planning phase.
- Capability inventory: The agent has the ability to write files, perform web research (
SKILL.md), and execute local git commands such as branching and committing (references/PLANNING_PROCESS.mdstep 15). - Sanitization: No sanitization or escaping mechanisms are specified for the external data before it is used in generated artifacts or local commit messages.
Audit Metadata