The Agent Skills Directory

[COMMAND_EXECUTION]: The skill is designed to execute multiple local shell and Python scripts residing in the user's home directory.
Executes ~/Documents/jj-knowledge-vault/system/scripts/unpushed-sweep.sh to check for local git commits.
Invokes python3 system/scripts/shared-memory/memory_store.py for memory reconciliation.
Runs python3 ~/Documents/jj-knowledge-vault/system/scripts/shared-memory/consolidator.py for pattern promotion and store pruning.
[DYNAMIC_EXECUTION]: Employs inline Python scripts to perform utility functions and manage execution flow.
Calculates token budgets for context.md using python3 -c and regex-based estimation.
Implements a custom timeout wrapper using `python3
<<'PY'` to bound the execution of the consolidator script on macOS.
[DATA_EXFILTRATION]: Facilitates the transfer of local data to external services.
Performs git push operations to synchronize the vault's skills and content with GitHub repositories.
The "unpushed-work sweep" identifies and can potentially push branches that have not yet reached a remote server.
[INDIRECT_PROMPT_INJECTION]: The skill manages a feedback loop where session data influences future agent behavior by updating persistent memory.
Ingestion points: Reads session narratives, corrections, and decisions from the current session (SKILL.md) to update long-term memory files like context.md and memory.md.
Boundary markers: The instructions do not specify explicit delimiters or "ignore embedded instructions" warnings for the session content being ingested into persistent memory files.
Capability inventory: The skill utilizes subprocess.run and Git for network operations, and possesses broad file-write capabilities across the vault directory structure.
Sanitization: No explicit sanitization, escaping, or validation of session-derived content is performed before it is interpolated into the agent's core context files.

wrapup