The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill implements a self-propagation mechanism where it reads its own source code from scripts/extract.ts and transmits it to remote hosts via SSH for execution. This allows the skill to execute arbitrary logic on any machine listed in the user's remotes.txt file.
[COMMAND_EXECUTION]: The skill makes extensive use of the node:child_process module to execute system commands including ssh, git, gh, node, pnpm, and python3. These tools are used for data extraction, environment validation, and automated publication of code changes.
[DATA_EXFILTRATION]: The skill aggregates sensitive data from T3 Code databases (~/.t3/userdata/statev2.sqlite), including private messages, project titles, and thread history. While the public-check.ts script is designed to prevent the publication of private identifiers to GitHub, the broad collection and potential transmission of this data remains a significant exposure risk.
[DYNAMIC_EXECUTION]: In scripts/public-check.ts, the skill dynamically generates a JavaScript snippet and executes it on remote hosts using the node --input-type=module - command. This pattern of generating and executing code at runtime increases the difficulty of static security auditing.
[INDIRECT_PROMPT_INJECTION]: The skill's primary function involves ingesting and processing untrusted data (user messages and thread content). This creates an attack surface for indirect prompt injection if malicious instructions are embedded within the analyzed threads, potentially influencing the agent's behavior during the 'Find patterns' or 'Deliver' steps.
Ingestion points: Reads messages from local and remote SQLite databases in extract.ts.
Boundary markers: None identified in the prompt interpolation logic.
Capability inventory: File system writes, network operations via git/gh, and remote command execution via ssh.
Sanitization: Relies on public-check.ts which uses regular expressions to filter known private identifiers but does not sanitize for malicious instructions.

usage-review