The Agent Skills Directory

[SAFE]: The skill's operation is confined to local file reading and writing within a specified artifact directory. It includes a mandatory 'HARD-GATE' constraint that explicitly forbids writing code, creating files outside the artifact folder, or taking any implementation action.
[PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface due to its requirement to ingest content from external files like goal-definition.md and context-map.md. 1. Ingestion points: SKILL.md (Step 1 and Step 2). 2. Boundary markers: Absent; files are read into the context without explicit delimiters or instructions to ignore embedded commands. 3. Capability inventory: The skill is restricted to generating markdown documentation; it lacks network access, command execution, or dynamic code loading capabilities. 4. Sanitization: No sanitization is performed on the ingested file content.

brainstorming