The Agent Skills Directory

[COMMAND_EXECUTION]: The checkpoint.py script executes the git command-line tool using the subprocess module to retrieve repository metadata, including commit hashes, messages, file status changes, and line-level statistics.
[PROMPT_INJECTION]: The skill is subject to indirect prompt injection risks because it processes data from external sources, such as git commit messages and user-influenced CLI logs, which are then passed to a subagent for pattern analysis.
Ingestion points: Data is ingested from git log output, the cli-tools.jsonl log file, and various teammate work logs located in .claude/logs/agent-teams/.
Boundary markers: The script organizes the collected data into structured markdown sections with headers and code blocks before presenting it to the subagent for analysis.
Capability inventory: The subagent analysis can lead to the creation of new skill definitions in the .claude/skills/ directory, though the documentation indicates this requires explicit user approval.
Sanitization: No explicit sanitization or filtering of the ingested log content is performed before it is interpolated into the analysis prompt.

checkpointing