plan-summary-md
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted documents provided by the user to generate summaries, creating a risk that malicious instructions inside those documents could influence the agent's behavior.
- Ingestion points: The
scripts/collect_plan_evidence.pyscript reads user-specified files with.md,.markdown, or.txtextensions. - Boundary markers: Detailed workflow instructions in
references/plan-summary-workflow.mdexplicitly direct the agent to treat all ingested content as untrusted data and to ignore any prompt-like text as instructions. - Capability inventory: The skill has the ability to read local text files (via
collect_plan_evidence.py) and write summary artifacts to the.plan-summariesdirectory (viagenerate_plan_summary.py). - Sanitization: Input paths are rigorously validated for control characters and protocol markers. The output generator script employs HTML escaping and JSON-safe encoding to prevent injection in the generated reports.
Audit Metadata