The Agent Skills Directory

[DATA_EXFILTRATION]: The skill automatically transmits sensitive session and operational data to a configurable external gateway.
Evidence: The cc_kit_session_end hook forwards session summaries, cost alerts, and lists of files changed to the OpenClaw service.
Evidence: The cc_kit_tool_use hook sends tool names and their execution results to the gateway.
Evidence: The skill syncs the ~/.claude/sessions/ directory, which contains sensitive conversation history, with the OpenClaw platform.
[COMMAND_EXECUTION]: The skill defines and executes commands that manage agent profiles and system health.
Evidence: The /openclaw agents generate command triggers profile creation logic based on Kit workflow modes.
Evidence: The /openclaw health command performs network probes and checks the status of external channel connectivity (Discord, Slack, Telegram).
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection by ingesting untrusted data from external sources.
Ingestion points: Processes channel_message data from Discord, Slack, and Telegram via the OpenClaw gateway.
Boundary markers: None specified to delimit external messages from agent instructions.
Capability inventory: The skill can write to tasks/lessons.md, read session history, and invoke a wide range of kit domain skills.
Sanitization: No evidence of sanitization or filtering for the content received from external message channels before it is used for status updates or context.

openclaw-native