The Agent Skills Directory

[PROMPT_INJECTION]: The skill includes adversarial prompt fixtures in tests/fixtures/adversarial.md specifically designed for robustness testing. These fixtures demonstrate how the agent should identify and resist manipulative user input that attempts to bypass safety warnings or force overconfident approvals, improving overall security posture.\n- [COMMAND_EXECUTION]: The script scripts/quick_validate_skill.py utilizes subprocess.run() to execute the local test contract in tests/test_skill_contract.py. This is a standard administrative function for verifying the skill's integrity during development and deployment.\n- [REMOTE_CODE_EXECUTION]: The file tests/test_skill_contract.py uses the Python compile() function to perform syntax validation on scripts within the skill directory. This dynamic execution is strictly scoped to local syntax checking for maintenance purposes.\n- [EXTERNAL_DOWNLOADS]: The skill references several external research papers, documentation, and asset files from trusted domains such as Microsoft, Anaconda, arXiv, and the author's official GitHub repository. These references are used for grounding and visual presentation, following established safe-source practices.

cost-aware-compute-planner