The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits a surface area for indirect prompt injection as it ingests untrusted data from multiple sources to influence its output and implementation actions.
Ingestion points: Processes user input via interview phases, reads code repository content (files, history, tests), and scans docs/wiki/ for project context.
Boundary markers: While the instructions include a 'design confirmation' step to expose understanding bias, there are no explicit delimiters or system-level instructions to ignore embedded instructions within the ingested wiki or code files.
Capability inventory: The skill can write documentation files to docs/scope/ and invoke the do-scoped implementation skill to perform code modifications based on gathered requirements.
Sanitization: No explicit sanitization, filtering, or escaping of external content is described before it is interpolated into implementation contracts or specifications.

scope