pm-skills

Pass

Audited by Gen Agent Trust Hub on Sep 17, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes potentially untrusted data from Jira issues. * Ingestion points: Data is pulled using the Atlassian MCP and stored in snapshot.json before being normalized by scripts/jira_snapshot_bridge.py. * Boundary markers: The skill follows a structured 'Observe-Choose-Act-Verify' loop described in references/pm_loop_playbook.md, though explicit prompt delimiters for data interpolation are not defined. * Capability inventory: The skill executes local Python scripts and leverages Atlassian MCP tools to manage issues and transitions. * Sanitization: The bridge scripts normalize structured fields but do not perform sanitization on natural language fields like summaries or descriptions.
  • [EXTERNAL_DOWNLOADS]: The skill connects to the well-known Atlassian Remote MCP at https://mcp.atlassian.com/v1/sse for data synchronization and project resource access.
  • [COMMAND_EXECUTION]: The orchestrator runs several bundled Python scripts to manage routing (pm_goal_router.py), data processing (jira_snapshot_bridge.py), and governance enforcement (delivery_loop_gate.py).
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 17, 2026, 12:37 AM
Security Audit — agent-trust-hub — pm-skills