closing-obsolete-issues
Pass
Audited by Gen Agent Trust Hub on Jun 20, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: Indirect prompt injection surface identified.
- Ingestion points: The script
scripts/fetch_issue_details.shfetches untrusted issue descriptions and comments from the publicflutter/devtoolsrepository via the GitHub CLI. - Boundary markers: The script uses textual headers such as
Description:and--- ALL COMMENTS ---to structure the output, but it does not employ formal data encapsulation or escaping for the ingested content. - Capability inventory: The agent has the capability to modify the reference file
references/rationale_templates.mdand close GitHub issues using thegh issue closecommand. - Sanitization: There is no explicit sanitization or filtering of the issue content before it is processed by the agent.
- Mitigation: The skill instructions include a 'CRITICAL MANDATE' requiring the agent to present all candidate issues and drafted comments to the user for explicit approval before any state-changing actions are performed.
Audit Metadata