The Agent Skills Directory

[SAFE]: The skill's primary functionality is managing session state within the local environment, which is a standard workflow enhancement.- [INDIRECT_PROMPT_INJECTION]: The skill ingests historical session data and conversation trailers, creating an attack surface for indirect instruction injection.\n
Ingestion points: The skill reads existing CONTEXT-*-llm.md files from the local filesystem and parses  trailers from the current message transcript.\n
Boundary markers: The skill uses structured HTML comments and specific markdown section headers to delimit and organize processed data.\n
Capability inventory: The skill is authorized to use Bash, Read, Write, and Edit tools for file management and execution.\n
Sanitization: The skill performs a mandatory secret scrubbing pass to identify and replace PII, API keys, and tokens with redacted placeholders before storage.- [COMMAND_EXECUTION]: The skill executes shell commands via the Bash tool to list directories and archive files. Security is maintained through strict regex validation of the stream argument (^[a-zA-Z0-9_-]{1,50}$), which prevents unauthorized command or path manipulation.

save-context