report-to-keep
Pass
Audited by Gen Agent Trust Hub on Oct 13, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONPRIVILEGE_ESCALATIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes report content which represents an ingestion point for external data.
- Ingestion points: Ephemeral reports are read via OKF-MCP as described in SKILL.md.
- Boundary markers: The skill relies on OKF validation and Loom report profile compliance to establish provenance rather than textual delimiters.
- Capability inventory: The skill utilizes the loom_report_promote tool to perform file writes to the repository.
- Sanitization: Instructions in ASSESSMENT.md and QA.md require verification of report integrity and provenance before promotion, minimizing the risk of processing malicious instructions.
- [PRIVILEGE_ESCALATION]: The skill includes explicit instructions to prevent directory traversal and symlink escapes during file operations.
- Evidence: The QA contract in QA.md requires blocking path traversal and alternate-path tricks to ensure operations are confined to the ephemeral-reports/ and docs/reports/ directories.
- [COMMAND_EXECUTION]: The skill utilizes designated tools for discovery and promotion tasks to avoid unsafe shell access.
- Evidence: SKILL.md mandates the use of OKF-MCP for report discovery and loom_report_promote for file movement, rather than generic shell commands.
Audit Metadata