The Agent Skills Directory

[COMMAND_EXECUTION]: The skill invokes local management scripts (such as plan-manager.py, learning-db.py, and feature-state.py) and Git utilities to maintain task state and session context. These operations use predictable local paths in ~/.claude/scripts/ and the project root.
[PROMPT_INJECTION]: The skill processes data from several ingestion points, including task_plan.md, HANDOFF.json, and .continue-here.md, to reconstruct session state. This creates an indirect prompt injection surface as there are no explicit boundary markers or sanitization routines for the content of these files. However, this data processing is necessary for the skill's core purpose of cross-session context preservation.
[DATA_EXFILTRATION]: All data operations are restricted to the local filesystem and the current Git repository. No network-based exfiltration patterns or requests to non-whitelisted domains were found.

planning