execution-trace
Pass
Audited by Gen Agent Trust Hub on May 18, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill implements a logical workflow for project management and traceability. It maintains data integrity by instructing the agent to verify evidence before claiming task completion and to avoid inventing identifiers for tickets, PRs, or commits.\n- [SAFE]: Analysis of the indirect prompt injection surface as per Category 8 guidelines:\n
- Ingestion points: Reads project requirements from
.agents/requirements.mdand project context from.agents/project-context.md. It also retrieves external data from trackers (Jira, Linear) and repositories (GitHub, GitLab) such as ticket and PR descriptions.\n - Boundary markers: None explicitly defined in the instructions to separate external data from the prompt context.\n
- Capability inventory: The skill is permitted to write to
.agents/execution-trace.mdand log local command outputs.\n - Sanitization: No explicit sanitization of external content is requested; however, the structured nature of the output (tables) limits the impact of embedded instructions.\n- [SAFE]: The skill accesses project metadata but restricts output to the local workspace within the
.agents/directory. No unauthorized network exfiltration or credential exposure patterns were identified.
Audit Metadata