rca
Pass
Audited by Gen Agent Trust Hub on Jul 29, 2026
Risk Level: SAFE
Full Analysis
- [PROMPT_INJECTION]: The skill processes untrusted data from external sources including Linear issue descriptions, Slack threads, and Git logs, creating an indirect prompt injection surface. This risk is mitigated by a mandatory human-in-the-loop requirement, where the agent must draft the analysis for user review and obtain explicit approval before posting to Linear.
- Ingestion points: Linear issues, Slack threads, Git logs, GitHub PR bodies, and SQL results via devsql.
- Boundary markers: None explicitly defined in the instructions to separate untrusted data.
- Capability inventory: Tool access includes linear, git, gh, and devsql via Bash, allowing for both data reading and external posting.
- Sanitization: No programmatic sanitization is specified, relying instead on manual user review of the drafted content.
- [COMMAND_EXECUTION]: The skill employs Bash tools scoped to linear, git, gh, and devsql. These tools are utilized for legitimate bug investigation tasks such as viewing issue details, searching code history, and querying databases.
- [DATA_EXFILTRATION]: Data is retrieved from internal environments and eventually posted to an external platform. This process is controlled by a strict approval flow, ensuring no data is exfiltrated without direct user confirmation.
Audit Metadata