The Agent Skills Directory

[COMMAND_EXECUTION]: The Python implementation playbook uses the subprocess.run method to execute git commands (e.g., git diff --stat, git diff --name-status) to analyze project changes.
Evidence: cmd = f"git diff --name-status {base_branch}...HEAD" and result = subprocess.run(cmd.split(), ...) in resources/implementation-playbook.md.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from git diffs to generate summaries and checklists. A malicious actor could embed instructions in the code changes designed to manipulate the agent's PR summary or checklist generation.
Ingestion points: The PRAnalyzer class in resources/implementation-playbook.md reads file names and diff statistics via git diff.
Boundary markers: The instructions lack specific delimiters or "ignore embedded instructions" prompts when processing the raw diff content.
Capability inventory: The skill has the capability to execute shell commands via subprocess.run and write to the agent's output buffer.
Sanitization: The provided Python snippets perform regex parsing and string splitting but do not include explicit sanitization of the content of the diff itself before it is processed by the LLM.

comprehensive-review-pr-enhance