The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to run local TypeScript scripts, specifically MigrateScan.ts and MigrateApprove.ts, using the bun runtime. These scripts are located in the user's home directory (~/.claude/LIFEOS/TOOLS/), indicating they are part of the local environment's tooling.
[DATA_EXFILTRATION]: The skill includes a mandatory step to send status notifications via a curl POST request to http://localhost:31337/notify. Because this request is restricted to localhost, it does not transmit data to external servers or represent a risk of exfiltration.
[PROMPT_INJECTION]: The skill processes untrusted data from external sources such as .md files, directory scans, and stdin. This creates an indirect prompt injection surface where malicious instructions embedded in the source content could attempt to influence the agent during classification or review.
Ingestion points: File paths, directory recursion, and piped stdin content (Phase 1).
Boundary markers: The skill applies HTML comments to committed content to maintain provenance (Rules section).
Capability inventory: Shell command execution via bun, local network requests via curl, and file system writes to TELOS/ and MEMORY/ directories.
Sanitization: No specific content sanitization or instruction filtering is described for the ingestion phase, though the skill relies on confidence scores and user approval loops to mitigate classification errors.

Migrate