The Agent Skills Directory

[DYNAMIC_CONTEXT_INJECTION]: The skill uses the !command syntax in Phase 0.0 to execute a shell script at load time: !`${CLAUDE_PLUGIN_ROOT}/scripts/discover-config.sh. This executes external logic automatically when the skill is accessed, bypassing standard agent interaction cycles.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data that could contain malicious instructions.
Ingestion points: Reads git commit history (Phase 1.1) and issue bodies from backends (Phase 3.2).
Boundary markers: None specified for handling external text.
Capability inventory: File read/write, bash shell execution, and git command operations.
Sanitization: No evidence of input validation or escaping for the ingested issue descriptions or commit messages used in decision-making logic.
[COMMAND_EXECUTION]: The skill relies on complex shell operations and the yq utility to modify project state. It also references external logic in references/reconcile-<backend>.md files which are not provided for analysis, making the full scope of potential command execution unverifiable.

reconcile