The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses wc -l and wc -c to monitor the size of HANDOFF.md to ensure it stays within a target size (150-200 lines). These are benign, read-only commands used for file size discipline.
[DATA_EXPOSURE]: The skill manages files that may contain project metadata, but it operates entirely within the local project structure (.claude/throughline/, .agent/handoff/) and emphasizes user privacy by confirming the scope before reading logs. There are no network operations or exfiltration patterns.
[PROMPT_INJECTION]: While the skill involves reading and processing session logs (which could contain untrusted data), it implements a 'human gate' where no changes are applied to the system without explicit user review and approval of each candidate lesson.
[SAFE]: The skill follows least-privilege principles by proposing issues for external repositories rather than attempting to edit them directly.

consolidate