The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local Python scripts (e.g., scripts/run_eval.py or eval/run.py) via the bash tool. While this is the primary function of the skill, it involves running arbitrary code located within the user's project environment.
[DYNAMIC_EXECUTION]: In Phase 1 and 3, the skill dynamically discovers the path to an evaluation script and subsequently executes it using the Python interpreter. The execution of code from dynamically determined paths introduces a level of uncertainty regarding the exact instructions being run.
[INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection through the processing of untrusted data.
Ingestion points: The agent reads content from gold_standards/, test_data/, and generated results/ (referenced in SKILL.md Phases 5 and 6).
Boundary markers: The instructions do not specify the use of delimiters or 'ignore' instructions when processing external text data.
Capability inventory: The skill utilizes bash to run commands and write to generate evaluation reports and suggestions.
Sanitization: There is no evidence of sanitization or filtering of the inputs being analyzed before they are used to generate prompt improvement recommendations.

sc-evaluate