The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes standard version control commands including git status, git commit, and git push to manage the transition of handoff documents between local and remote environments.\n- [DATA_EXFILTRATION]: When the cloud option is invoked, the skill is designed to transmit conversation summaries to a remote GitHub repository. This behavior is the primary function of the skill for portability and includes safeguards such as requiring explicit user approval for every push operation.\n- [INDIRECT_PROMPT_INJECTION]: As the skill ingests conversation history to generate summaries for a future agent, it possesses a vulnerability surface common to all multi-session agents. A malicious user or external data source in the current session could attempt to inject instructions intended to influence the behavior of the agent in the next session.\n
Ingestion points: Reads the current conversation context (history) to generate the summary.\n
Boundary markers: The instructions mandate a specific Mermaid diagram and "Suggested skills" structure, providing some structural differentiation, but no explicit "ignore" delimiters are defined for the next agent's ingestion.\n
Capability inventory: File writing (local temp and repository paths) and Git operations (commit, push).\n
Sanitization: Includes mandatory instructions to redact API keys, tokens, passwords, and personal data from the generated output.\n- [EXTERNAL_DOWNLOADS]: The documentation references external resources and templates from github.com/mattpocock/skills for vendoring purposes.

handoff