The Agent Skills Directory

[SAFE]: The skill executes local commands through the skillwiki utility, including path for vault resolution and validate for frontmatter verification. These operations are restricted to managing local vault metadata.
[DATA_EXFILTRATION]: The skill reads project-specific documents and checks system-level configurations such as crontab to audit and verify completion of tasks. These read operations are limited to the local environment and intended for task validation.
[PROMPT_INJECTION]: The skill processes user-generated or previously written project files, presenting a potential surface for indirect prompt injection.
Ingestion points: Reads spec.md, tasks.md, and log.md from within projects/{slug}/work/ directories.
Boundary markers: None identified; the instructions do not specify the use of delimiters when reading file content.
Capability inventory: Filesystem read/write access, skillwiki command execution, and inspection of system configuration files.
Sanitization: File content is ingested for state verification without explicit sanitization or filtering logic.

proj-work