The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructions in SKILL.md suggest running a local Python script python3 scripts/lint_goal.py - to validate goal contracts. This is a standard linting operation and is limited to checking string length and specific placeholder tokens.
[NO_CODE]: The provided Python script scripts/lint_goal.py performs safe string manipulations using regular expressions to detect common placeholders like 'TODO' or 'TBD'. It does not perform network operations, access sensitive file paths, or use unsafe execution functions like eval() or exec().
[DATA_EXPOSURE]: No hardcoded credentials, API keys, or access to sensitive local directories (e.g., .ssh, .aws) were found. The skill operates solely on the goal text provided by the user.
[PROMPT_INJECTION]: The instructions focus on defining clear boundaries and acceptance criteria for tasks. There are no attempts to bypass safety filters or override system-level constraints. The references/cdc-case.md file explicitly advises against copying benchmark-specific restrictions that might bias research, promoting neutral and accurate goal drafting.

goal-writer