The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests and processes user-supplied session logs in JSONL format to identify toil candidates.
Ingestion points: scripts/recent_human.py (line 123) reads raw data from session files provided via command-line arguments.
Boundary markers: The script uses # My request for Codex: as a marker to isolate user requests and applies regex patterns in MACHINE_ECHO_PATTERNS to filter out machine-generated content.
Capability inventory: The skill uses the extracted data to score and rank tasks, and it has the capability to write output reports to the .agents/scratch/toil-mining/ directory.
Sanitization: The script normalizes whitespace and line endings but does not sanitize the extracted text for malicious instructions that might target the agent's behavior during the analysis phase.
[COMMAND_EXECUTION]: The test suite utilizes shell command execution to verify the functionality of the extraction script.
Evidence: tests/test_recent_human.py (line 39) uses subprocess.run to call the Python interpreter and execute scripts/recent_human.py with specific test parameters. This is limited to the testing environment and uses programmatically defined arguments.

toil-mining