The Agent Skills Directory

[DATA_EXFILTRATION]: The skill automatically extracts sensitive authentication tokens and session cookies from the user's local browser storage to authenticate with social media platforms.
scripts/lib/chrome_cookies.py: Accesses the macOS Keychain using security find-generic-password to retrieve the Chrome Safe Storage encryption key and utilizes openssl to decrypt private cookie values.
scripts/lib/cookie_extract.py: Programmatically scans the file system for cookies.sqlite (Firefox) and Cookies.binarycookies (Safari) to harvest auth_token, ct0, and _session_id credentials.
scripts/lib/env.py: Manages and aggregates these harvested secrets for use in outbound API requests to X/Twitter and Truth Social.
[COMMAND_EXECUTION]: The skill relies on executing various local binaries and shell commands for its core functionality and setup.
scripts/lib/github.py: Executes the gh CLI tool to retrieve active GitHub authentication tokens.
scripts/lib/youtube_yt.py: Runs the yt-dlp binary to search for videos and extract transcript data.
scripts/lib/subproc.py: Implements a wrapper around subprocess.Popen to execute arbitrary command lists with process group management.
scripts/lib/xurl_x.py: Shells out to the xurl CLI tool for official X API integration.
[EXTERNAL_DOWNLOADS]: The skill manages software installations at runtime and communicates with many third-party services.
scripts/lib/setup_wizard.py: Automatically triggers the installation of system packages using brew install yt-dlp during its first-run configuration.
scripts/lib/apify.py: Executes remote scraping logic via Apify's run-sync-get-dataset-items API for TikTok, Instagram, and Threads.
scripts/lib/http.py: Centralizes network operations to numerous external domains including ScrapeCreators, Algolia, and various LLM providers.
[PROMPT_INJECTION]: The skill instructions attempt to override system tool behavior and process untrusted data.
SKILL.md: Contains a 'VOICE CONTRACT' that explicitly instructs the AI to 'SUPERSEDE' and 'IGNORE' the mandatory formatting and citation requirements of the system's core WebSearch tool.
scripts/lib/rerank.py: Ingests and processes large volumes of untrusted content from social platforms, which represents a significant attack surface for indirect prompt injection.

last30days