implement-issues
Pass
Audited by Gen Agent Trust Hub on Sep 17, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from GitHub issues without sanitization.
- Ingestion points: The
assets/issue-brief.mdtemplate extracts the<full issue body, verbatim>and provides it to the subagent. - Boundary markers: Missing clear delimiters or explicit instructions to discard instructions found within the issue body.
- Capability inventory: The agent can modify code and execute arbitrary shell commands via project-defined test runners.
- Sanitization: There is no escaping or filtering applied to the ingested issue text.
- [COMMAND_EXECUTION]: The skill automatically runs shell commands found in the repository metadata.
- Evidence: Stage 4 in
SKILL.mdperforms automated verification by running 'the project's test, lint, and typecheck commands (from AGENTS.md or the project's manifest)'. If these configuration files are manipulated by a malicious actor or an injected subagent, the dispatcher will execute the resulting commands.
Audit Metadata