The Agent Skills Directory

[COMMAND_EXECUTION]: The orchestration script scripts/daily_releases.py and the test script scripts/test_daily_releases.py use subprocess.run to invoke other bundled Python scripts such as list_daily_ranges.py and collect_day_dataset.py. These calls are made using the current Python interpreter (sys.executable) to manage the release pipeline.
[INDIRECT_PROMPT_INJECTION]: The skill processes external, potentially untrusted data including git commit subjects, bodies, and diffs. This information is interpolated into prompts for AI subagents as documented in references/synthesis_prompt.md and notes/ai-analysis-subagent-fix.md.
Ingestion points: Commit metadata and GitHub issue records are collected in scripts/collect_day_dataset.py, while diffs are processed in scripts/bucket_day_data.py.
Boundary markers: The prompt templates in references/synthesis_prompt.md utilize <content> tags to separate bucket data from the core instructions.
Capability inventory: The agent has permissions to execute the pipeline's Python scripts, write artifacts to the local file system, and interact with the GitHub API to manage tags and releases.
Sanitization: There is no evidence of sanitization or character escaping performed on the raw commit data or diffs before they are embedded into the LLM context.
[DATA_EXFILTRATION]: The skill utilizes a GITHUB_TOKEN to retrieve repository data (commits, diffs, issues). While this is the primary function of the tool, it involves handling sensitive authentication material and broad repository content.
[SAFE]: The utility scripts/daily_releases_lib/daily_releases_lib/github_utils.py provides an option to disable SSL verification via the GITHUB_SSL_VERIFY environment variable. This is documented as a configuration option for specific environment needs rather than a hardcoded security bypass.

daily-releases