The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/tapi-auth.js modifies several shell initialization files to inject environment variables.
Affected files: ~/.bashrc, ~/.zshenv, ~/.profile, ~/.zprofile, ~/.config/fish/config.fish, and PowerShell profiles.
These modifications alter the shell environment every time a new session is started, establishing a persistence mechanism on the host system.
[DATA_EXFILTRATION]: The setup process involves reading from and writing to sensitive configuration paths on the host system.
Targeted paths: ~/.openclaw/openclaw.json, ~/.clawdbot/moltbot.json, and multiple shell resource files.
The script attempts to automatically backup and rewrite these files, which can expose the agent's internal configuration and user environment variables.
[PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection by processing untrusted data from an external API.
Ingestion points: Video transcripts fetched from https://transcriptapi.com/api/v2/youtube/transcript are processed in the main skill flow.
Boundary markers: None. The skill does not instruct the agent to use delimiters or specific ignore-instructions for the transcript content.
Capability inventory: Shell command execution (curl) and system file-write capabilities via the authentication script.
Sanitization: None. Transcripts are ingested directly for analysis and summarization without validation, escaping, or filtering of embedded instructions.

video-transcript