ns-project-manager
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [COMMAND_EXECUTION]: The agent is instructed to execute local Python scripts (rice_wsjf.py, dag_order.py, pert_montecarlo.py, flow_metrics.py, okr_progress.py) to perform RICE/WSJF scoring, DAG layering, Monte Carlo simulations, and flow metrics calculations.
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data such as meeting transcripts, Slack messages, and emails which could contain malicious instructions.
- Ingestion points: Untrusted data enters the context via transcripts and informal communications (documented in SKILL.md, references/07-meeting-digest.md, and references/09-nl-to-workflow.md).
- Boundary markers: The skill uses [ASSUMPTION] tags and explicit '⚠️ Requires human review' warnings in its output.
- Capability inventory: The skill can write Markdown artifacts to the docs/ folder and mutate the environment by creating or updating GitLab issues via MCP.
- Sanitization: The skill mitigates risks by applying strict structured data frameworks (INVEST, Gherkin, WBS) to normalize all processed requirements.
- [EXTERNAL_DOWNLOADS]: The skill fetches current currency exchange data from the Banco Central do Brasil to calculate budget estimates in local currency. This uses a well-known public financial service.
Audit Metadata