The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a local shell script (larch.sh) using the Bash tool. It passes user-supplied arguments directly as flags to the CLI command, which is a standard pattern for diagnostic tools but requires the agent to handle argument safely.
[INDIRECT_PROMPT_INJECTION]: The skill processes potentially untrusted external data from multiple sources including JSON, TSV, and JSONL files.
- Ingestion points: The analyzer joins data from difficulty-rating.json, findings-classification.tsv, review-findings-full.jsonl, review-findings.ndjson, and rejected-analysis/verdicts.tsv found in the repository.
- Boundary markers: There are no explicit instructions or delimiters mentioned to separate or ignore instructions that might be embedded within the logs or artifacts being analyzed.
- Capability inventory: The skill utilizes the Bash tool to execute a script and the Read tool to access files.
- Sanitization: No explicit sanitization or validation of the input log content is described before the output is relayed to the user.

difficulty-calibration