The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses several shell scripts to execute local system commands for its operation. These include git for repository management and session history extraction, grep, sed, and awk for text processing and basic JSON parsing, and mkdir for directory management. These operations are consistent with the skill's purpose of managing a session-end lifecycle.
[DATA_EXFILTRATION]: The skill transmits data to a remote Git repository (team-ai-directives) as part of its core functionality. It uses git push to upload CDR drafts and session reports to an orphan branch. This network operation is the intended method for publishing team-wide learnings and directives, with the destination repository being configurable by the user or project settings.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted session data, including git diff outputs and logs, to identify reusable patterns. While this data could theoretically contain malicious instructions designed to influence the agent, the skill mitigates this risk through a mandatory interactive 'Batch Review' phase. This human-in-the-loop requirement ensures that no extracted patterns are accepted or published without explicit user approval.
[SAFE]: The skill incorporates a dedicated privacy-scrubbing mechanism (scripts/scrub-session.sh) to protect user data. This script uses regular expressions to identify and redact common sensitive patterns, such as API keys, authentication tokens, and passwords, from session summaries before they are saved or shared.

team-levelup