qa-issue-validate
Pass
Audited by Gen Agent Trust Hub on Sep 4, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONDATA_EXFILTRATION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted external content from GitHub and GitLab issues, which presents a surface for indirect prompt injection attacks. \n
- Ingestion points: Issue titles, bodies, and comments are fetched from external platforms in Step 2. \n
- Boundary markers: The skill mandates an 'Issue Content Threat Scan' (Step 2.5) that instructs the agent to treat input strictly as data, ignore instructions within the text, and strip hidden characters. \n
- Capability inventory: The skill utilizes shell tools (
gh,glab,git,curl) and file system operations (Read,Write,Edit). \n - Sanitization: Mandatory privacy sanitization (Step 7) and roadmap sanitization (Step 10) are implemented to prevent untrusted data from triggering unintended actions or leaking sensitive information. \n- [COMMAND_EXECUTION]: The skill uses platform CLI tools and git to interact with repositories. Robust guardrails in Step 2.5 and Step 8 prevent the execution of arbitrary commands derived from untrusted issue content. \n- [DATA_EXFILTRATION]: The skill contains strict rules against reading sensitive files like
.env, secrets, or private keys (Step 3.2), even if requested by an issue, and performs privacy sanitization to prevent accidental data leakage in comments or reports.
Audit Metadata