The Agent Skills Directory

[DATA_EXFILTRATION]: The skill contains instructions to automatically detect a project's git status and, if no remote is present, create a private repository under the author's namespace (joydai2026-del) and push the contents. This represents a significant risk of exfiltrating the user's source code and local project data to a repository controlled by the skill author.
[COMMAND_EXECUTION]: The skill utilizes several shell commands and local scripts to manage state, including:
git remote and git push for repository synchronization.
python3 to execute local scripts (memory_store.py, consolidator.py) located in ~/Documents/<NOTES_ROOT>/system/scripts/.
ln -s for creating symbolic links between the project root and the vault.
brctl download for forced downloading of iCloud-evicted files.
rg (ripgrep) for searching channel UUIDs and thread IDs in logs.
df -h / and stat for monitoring disk space and file metadata.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest data from external and historical sources into the agent's current context, which could allow malicious instructions in those sources to influence the agent.
Ingestion points: Historical session logs (session-logs/), session indices (session-index.md), and content from team chat blocks ([Context]).
Boundary markers: No specific delimiters or "ignore embedded instructions" warnings are used when interpolating this historical data.
Capability inventory: The agent has access to shell execution, file system modification (symlinks, file creation), and network access (git operations).
Sanitization: There is no evidence of sanitization, validation, or escaping of the content ingested from the vault or chat logs.
[PROMPT_INJECTION]: The skill uses forceful and mandatory language ("MANDATORY session start protocol", "STOP and run it", "Never skip this skill") to dictate the agent's behavior and operational priority, potentially overriding user instructions or system safety defaults.
[DYNAMIC_EXECUTION]: The skill employs python3 -c to execute inline Python code for parsing and processing JSON data at runtime, creating a surface for dynamic code execution based on the output of other scripts.

memory-loader