The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/create-pr-mr.sh executes several system commands including git, gh (GitHub CLI), jq, and curl to manage repositories and interact with APIs. These executions are handled safely using Bash arrays for gh arguments and jq --arg for building JSON payloads, which significantly mitigates shell injection risks.
[INDIRECT_PROMPT_INJECTION]: The skill processes external data that could contain malicious instructions.
Ingestion points: Data enters the skill via the --title, --body, and --body-file arguments in scripts/create-pr-mr.sh.
Boundary markers: The skill does not currently wrap the PR/MR description in explicit boundary markers or "ignore instructions" warnings to protect downstream agents that might read the created PR.
Capability inventory: The skill has capabilities to push code to repositories, create pull requests, and execute a sibling Node.js script for Jira linking.
Sanitization: Content is correctly escaped for API transit using jq, though the semantic content remains unfiltered for downstream consumers.
[DYNAMIC_EXECUTION]: The skill invokes a sibling script (link-pr-mr.js) using node at a path computed at runtime. The path is derived from the RHDH_JIRA_LINK_SCRIPTS environment variable or a relative path to the expected sibling directory ../../jira/rhdh-jira-link/scripts. This execution is a functional requirement for linking Jira issues and remains within the vendor's own skill ecosystem.

rhdh-pr-mr