The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The tools/docker-compose.yml file contains volume mounts that map the user's sensitive home directories for SSH (~/.ssh) and Git (~/.gitconfig) into the bot's Docker container. This grants the containerized environment read access to the user's private authentication keys and identity configuration.
[COMMAND_EXECUTION]: The /streak-telegram flow (Flow 13) automates the deployment of a persistent background service using docker-compose up -d. This involves executing shell commands that modify the host system's running processes and state.
[EXTERNAL_DOWNLOADS]: The tools/streak-bot.py and tools/streak-notify.py scripts communicate with the official Telegram Bot API (api.telegram.org) to send notifications and receive user commands.
[INDIRECT_PROMPT_INJECTION]: The Check-in flow (Flow 2, Step 3.5) includes an optional "Research" step where the agent is instructed to scan external sources for news and updates relevant to the user's challenge. This creates a potential vector for indirect prompt injection attacks.
Ingestion points: External news and updates retrieved during the Check-in planning phase (Flow 2).
Boundary markers: The skill does not define specific delimiters or instructions to ignore embedded commands in the research data.
Capability inventory: The skill performs filesystem writes (session notes, logs) and command execution (git/docker via tools).
Sanitization: No explicit sanitization or filtering of the retrieved research content is performed before processing.

streak