skills/pedronauck/skills/impeccable/Gen Agent Trust Hub

impeccable

Pass

Audited by Gen Agent Trust Hub on May 12, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTIONDATA_EXFILTRATIONEXTERNAL_DOWNLOADS
Full Analysis
  • [PROMPT_INJECTION]: The skill's 'Live' iteration workflow ingests HTML and text data from the user's browser, creating a surface for indirect prompt injection.
  • Ingestion points: Browser events received by live-server.mjs and processed in live-poll.mjs containing element.outerHTML and element.textContent.
  • Boundary markers: Present. The skill uses <!-- impeccable-variants-start --> delimiters and structural wrappers generated by live-wrap.mjs to isolate target content.
  • Capability inventory: The skill possesses extensive file system write capabilities (live-inject.mjs, pin.mjs, live-server.mjs) and authorized shell execution via Bash (npx commands).
  • Sanitization: Absent for content. While live-server.mjs validates the structure and types of incoming event payloads using regex, the ingested HTML/text content is not sanitized for instructions before being presented to the agent.
  • [COMMAND_EXECUTION]: The skill includes utility scripts that modify the agent's environment and project files to facilitate its workflow.
  • Shortcut Management: The pin.mjs script creates new SKILL.md files in harness-specific directories (such as .claude/skills/ or .cursor/skills/) to expose sub-commands as top-level agent commands.
  • File Injection: live-inject.mjs modifies project HTML files to insert script tags pointing to a local helper server.
  • [DATA_EXFILTRATION]: The skill starts a local HTTP server on 127.0.0.1 to receive visual context from the browser.
  • Local Server: live-server.mjs listens on a local port (defaulting to 8400) to receive screenshots, element data, and user annotations via POST requests.
  • Security Controls: The server generates a random UUID token required for all browser-to-server and agent-to-server communications.
  • [EXTERNAL_DOWNLOADS]: The skill references and downloads external resources from well-known and trusted providers.
  • Package Registry: The skill utilizes npx impeccable to run its core CLI logic, downloading the author's package from the official npm registry.
  • Media Assets: Design guidelines in reference/brand.md provide templates for fetching stock imagery from Unsplash.
Audit Metadata
Risk Level
SAFE
Analyzed
May 12, 2026, 05:53 PM