task-board-init
Pass
Audited by Gen Agent Trust Hub on Jul 25, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection during its analysis and 'doc sweep' phases. The orchestrator spawns subagents to ingest and classify untrusted data from existing repository documentation to populate the new task board.
- Ingestion points: Legacy documentation files identified in 'references/01-analysis.md' (e.g., TODO.md, ROADMAP.md, BACKLOG.md) and processed in 'references/06-doc-sweep.md'.
- Boundary markers: There are no explicit instructions for subagents to utilize delimiters or to ignore potential instructions embedded within the untrusted source text during the migration process.
- Capability inventory: The skill environment provides robust capabilities including the 'Bash' tool for shell operations, 'Write' and 'Edit' for file modifications, and the 'Task' tool for spawning autonomous subagents.
- Sanitization: The procedure does not specify validation or sanitization of the content extracted from the user's legacy files before it is re-authored into the generated Kanban artifacts.
- [DATA_EXFILTRATION]: The 'CLAUDE.md optimization' phase described in 'references/07-claude-md-optimize.md' explicitly scans for sensitive data including passwords, API keys, and private keys. While the intended purpose is to move these items to a gitignored local file for better security hygiene, the process involves an agent reading and identifying live credentials within the repository context, which is a sensitive operation.
Audit Metadata