The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses python3 -c to execute dynamic Python code blocks. Variables such as {JIRA_ID}, {branch_name}, and {pr_url} are interpolated directly into these Python strings within shell commands.
Evidence: Step 8 contains multiple instances like python3 -c "... client.transition_issue('{JIRA_ID}', 'In Review') ...".
Risk: If the values in CURRENT_TASK.md contain malicious characters (e.g., single quotes, semicolons), it could lead to arbitrary Python or shell code execution.
[PROMPT_INJECTION]: The skill relies on data extracted from CURRENT_TASK.md to drive its logic without implementing boundary markers or sanitization for the ingested content.
Ingestion point: CURRENT_TASK.md in Step 1 and Step 3.
Boundary markers: Absent.
Capability inventory: The skill has high-privilege capabilities including git push, gh pr merge, Jira API transitions, and file deletion (rm -f .claude/.ralph_loop_active).
Sanitization: None detected; the extracted JIRA_ID and branch_name are used directly in subsequent commands.
[COMMAND_EXECUTION]: The skill performs file system modifications and process management using shell commands.
Evidence: Step 11 executes rm -f .claude/.ralph_loop_active to deactivate the automation loop.
[EXTERNAL_DOWNLOADS]: The skill interacts with GitHub and Jira via CLI tools (gh) and custom Python integrations.
Evidence: Steps 5, 6, and 7 use git push and gh pr commands to interact with GitHub. Step 8 uses a Jira client to communicate with an Atlassian Jira instance.
Analysis: These operations target well-known services and are consistent with the skill's stated purpose.

finish-task