The Agent Skills Directory

[DATA_EXFILTRATION]: The skill is designed to read session history files located in ~/.claude/projects/, which contain sensitive logs of user interactions, source code, and tool outputs. This information is transmitted to an external service (MiniMax API) for processing.
[CREDENTIALS_UNSAFE]: The skill explicitly accesses a secret file at ~/.claude/.secrets/ccterrybot-telegram to retrieve the MINIMAX_API_KEY for authentication, exposing local credential storage practices.
[COMMAND_EXECUTION]: The skill uses the Agent tool to execute shell commands via bun to run a local TypeScript script (session-debrief.ts) with arguments derived from user input.
[PROMPT_INJECTION]: Indirect Prompt Injection Surface:
Ingestion points: Session histories are read from ~/.claude/projects/*.jsonl (SKILL.md).
Boundary markers: Absent; no delimiters or instructions to ignore embedded commands are used when processing logs.
Capability inventory: The skill has access to the Agent tool for shell execution and Bash for command running.
Sanitization: No sanitization or validation of the ingested session data is mentioned before it is processed or used in prompts.
[COMMAND_EXECUTION]: The skill contains self-modifying instructions under the 'Self-Evolving Skill' header, directing the agent to immediately edit the SKILL.md file to update its own instructions or practices based on runtime results.

session-debrief