The Agent Skills Directory

[Indirect Prompt Injection] (LOW): The skill ingests and synthesizes data from 'researcher' and 'scout' reports. This creates an attack surface where malicious instructions embedded in external research data could influence the final implementation plan produced by the agent. Evidence: 1. Ingestion points: 'researcher-XX-report.md' and 'scout-XX-report.md' files as defined in the plan directory structure. 2. Boundary markers: There are no explicit delimiters or instructions to ignore embedded commands in the ingested files. 3. Capability inventory: Writing documentation files to the local './plans/' directory. 4. Sanitization: No sanitization or validation of the ingested report content is performed.
[Data Exposure] (INFO): The skill is designed to read the local codebase and documentation (e.g., './docs/development-rules.md'). While it does not exfiltrate this information, it establishes a scope of access to potentially sensitive project intellectual property.

planning