iterate-research-questions
Pass
Audited by Gen Agent Trust Hub on Oct 5, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data which could contain malicious instructions.
- Ingestion points: Step 1 in
SKILL.mdidentifies multiple ingestion points, including user-authored comments, conversation history, ticket data, task collateral files, and changed repository evidence. - Boundary markers: The skill uses structural Markdown headers (e.g.,
## Retired Questions,## Revision log) to organize data, but it does not specify explicit delimiters (like XML tags or triple quotes) or specific "ignore embedded instructions" warnings for the content of the feedback items themselves. - Capability inventory: The skill possesses the capability to read files, execute shell commands (
ls -La), inspect repository contents, and write updates to the research-question artifacts. - Sanitization: Step 6 (
Validate the whole artifact) provides a validation layer that checks for YAML validity, Markdown structure, and identifier uniqueness, which helps prevent malformed content from breaking the artifact structure. - [COMMAND_EXECUTION]: The skill uses shell commands for environment inspection.
- Evidence: Step 1 in
SKILL.mdinstructs the agent to runls -La -- <task-directory>to resolve artifact locations and follow symlinked directories.
Audit Metadata