The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute local command-line tools such as grep and ripgrep (rg) for initial indexing and final completeness checks. It also mentions using project-specific tests or running the application as an automated 'oracle' to verify findings.
[DATA_EXFILTRATION]: The skill reads local codebase files and confirms evidence by re-opening files at specific line numbers. It stores its discovery state, including seen items and search angles, in a local journal file (quality-journals/work-exhaust-<task>.json) to support session resumption.
[PROMPT_INJECTION]: The skill processes untrusted content from the local codebase, which constitutes an indirect prompt injection surface.
Ingestion points: Search results from grep/rg and direct file content reads.
Boundary markers: None explicitly defined to isolate processed code content from agent instructions.
Capability inventory: Search commands (grep, rg) and the ability to run local tests or the application.
Sanitization: Findings are validated by re-reading the source file at the specified location (file:line) before being accepted.

work-exhaust