The Agent Skills Directory

[PROMPT_INJECTION]: The skill demonstrates an Indirect Prompt Injection surface by fetching and processing external, untrusted content to perform sensitive versioning tasks.
Ingestion points: PR titles, bodies, and code diffs are retrieved via gh pr view and gh pr diff in Phase 3 to be analyzed by a sub-agent.
Boundary markers: The prompt instructions for the classification sub-agent do not include explicit delimiters or instructions to ignore commands potentially embedded within the PR data.
Capability inventory: The skill has the capability to modify provider.yaml (which determines package versions) and changelog.rst, and it executes the breeze CLI tool for project orchestration.
Sanitization: There is no evidence of sanitization or filtering applied to the PR content before it is processed by the LLM.
[COMMAND_EXECUTION]: The skill relies on the execution of local shell commands and specialized tools to automate the release workflow.
It invokes git, gh, and the Airflow breeze CLI.
In Phase 5, the skill specifies the command prek run, which appears to be a typo for the standard pre-commit run tool used in the project. Executing misspelled or unknown binaries can be a security concern, although it is likely a non-malicious error in this specific context.

prepare-providers-documentation