The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill accepts an optional free-form hint from the user and forwards it as advisory context to the commit and pull request stages without sanitization.
Ingestion points: User-supplied [hint] parameter and {repo_root}/.gitkkal/config.json.
Boundary markers: Absent. The instructions do not define delimiters or specific instructions to ignore embedded commands within the hint or the configuration file.
Capability inventory: The skill (via its siblings) has the capability to execute git commit, git push, and gh (GitHub CLI) commands to create or update pull requests.
Sanitization: Absent. There is no validation or escaping of the input hint before it is interpolated into the workflow of the sibling skills.
[DYNAMIC_EXECUTION]: The skill implements a fallback mechanism that reads and follows instructions from other markdown files on the local file system.
Evidence: The skill attempts to read ../gitkkal-commit/SKILL.md and ../gitkkal-pr/SKILL.md to use as governing instructions if native tool invocation is not supported by the host. This pattern relies on the integrity of the local file system and the assumed presence of specific files in relative directories.

gitkkal-ship