project-delivery-perspective
Pass
Audited by Gen Agent Trust Hub on Oct 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill follows secure patterns by enforcing strict data boundaries between project constraints and quality facts. It explicitly prohibits the agent from altering status, execution results, or verdicts, which protects against social engineering or pressure to misrepresent project health.
- [INDIRECT_PROMPT_INJECTION]: The skill processes external data (e.g.,
delivery.jsonand user-supplied facts) which constitutes a vulnerability surface for indirect prompt injection. - Ingestion points: Data is ingested via external files (e.g.,
delivery.jsonin evaluation cases) and through user-supplied text prompts containing project facts. - Boundary markers: The skill defines a strict output contract with separate, mandatory sections for "Summary", "Project constraints", "Action tracking", and "Preserved quality facts" to prevent confusion between different data types.
- Capability inventory: The skill is limited to text analysis, formatting, and reporting. It has no capabilities for subprocess execution, file system modification, or network operations.
- Sanitization: Instructions mandate the preservation of facts exactly as supplied with their sources, and explicitly forbid the agent from interpreting or deciding on the data, which minimizes the impact of potential instructions embedded in the data.
Audit Metadata