The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses dynamic execution by instructing the agent to write a custom Python script (.scratch/split-timeline.py) and then execute it to process timeline data. It also executes complex Node.js one-liners to resolve network ports dynamically from the filesystem.
[DATA_EXFILTRATION]: The skill accesses sensitive configuration data by reading ~/.claude-mem/settings.json. While used for local configuration, this pattern involves reading files outside the active project workspace.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection through the ingestion of historical project data.
Ingestion points: The cm-timeline.md file (Step 2) contains the project's full history, including S## lines which store raw, unescaped prompts from previous user sessions.
Boundary markers: Absent. The prompt template provides a format key but does not include instructions to treat historical session content as untrusted or to ignore embedded instructions within those prompts.
Capability inventory: The skill possesses significant capabilities including shell command execution (node, git, curl), local file system manipulation (mkdir, mv), and the ability to generate and run new Python scripts.
Sanitization: Absent. The data is read via curl and passed into a subagent pipeline without filtering or escaping potentially malicious payloads in the history.

weekly-digests