The Agent Skills Directory

[SAFE]: The skill implements a structured workflow for decision-making and record-keeping without any evidence of malicious intent or dangerous behavior. All identified behaviors are consistent with its stated purpose.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface as it is instructed to read untrusted content from files within a project thread.
Ingestion points: The agent reads seed.md and decisions.md from docs/threads/<YYMMDDHHMMSSZ-slug>/ to establish context before starting the interview session.
Boundary markers: No explicit delimiters or specific warnings to ignore embedded instructions are defined for the ingested content.
Capability inventory: The skill's capabilities are limited to reading and appending to files within its designated thread directory. It possesses no tools for network exfiltration, system-level modifications, or subprocess execution.
Sanitization: No explicit sanitization or input validation is performed on the content read from the thread files.

discussion