The Agent Skills Directory

[COMMAND_EXECUTION]: The scripts/ci-watch.py script uses the Python subprocess.run module to execute the GitHub CLI (gh) and user-provided probe commands. The script follows security best practices by using argument vectors (argv) and disabling shell execution (shell=False), preventing shell injection vulnerabilities. This execution is the intended primary function of the CI monitoring tool.
[INDIRECT_PROMPT_INJECTION]: The ci-watch.py script ingests and processes JSON data from external command outputs (GitHub API and custom probes). The risk of the agent misinterpreting malicious content embedded in CI logs or job names is mitigated by the script's strict JSON schema validation and the skill's clear 'Analysis is not authorization' guardrails, which require explicit human confirmation for any downstream action.
[DYNAMIC_EXECUTION]: The documentation references the use of npx for installation and tool execution (e.g., npx tsx), as well as dynamic pipeline generation in systems like Buildkite. These patterns are standard for the described DevOps and optimization use cases and are presented with appropriate security warnings regarding untrusted code.

ci-cd-optimize