The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local shell commands to resolve paths and determine repository state.
Executes bash ~/.claude/coordination/resolve-coord-dir.sh to locate the coordination journal.
Executes git rev-parse --show-toplevel to identify the repository root.
[INDIRECT_PROMPT_INJECTION]: The skill ingests and processes untrusted data from multiple sources to perform its verification task.
Ingestion points: Processes user-pasted claims, repository summary files (docs/summaries/pause-*.md), and a shared coordination journal (journal.md).
Boundary markers: Absent; the instructions do not specify delimiters or warnings to ignore instructions embedded within the claims being verified.
Capability inventory: The agent can execute shell commands (bash, git), dispatch sub-agents (Agent tool), and append to files.
Sanitization: Absent; the skill passes extracted claims verbatim to the curator sub-agent.
[FILE_SYSTEM_MODIFICATION]: The skill prompts the user to append observational data to a log file located at ~/.claude/skill-observations/log.md.

curate