The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill launcher is designed to download and execute a native engine binary to the user's home directory (~/.impeccable/bin) on first run. This bypasses standard language-level sandboxing by executing compiled code.
[PERSISTENCE]: The skill's extended installation process (npx impeccable install) registers edit hooks. these hooks automatically trigger the skill's logic whenever UI files are modified, establishing a persistent execution presence in the development environment.
[EXTERNAL_DOWNLOADS]: The skill attempts to download approximately 854 MB of data from an external repository during installation and fetches additional binary components at runtime.
[DATA_EXFILTRATION]: The skill includes an anonymous usage telemetry feature that is enabled by default. While described as anonymous, this constitutes an automatic external data transmission of usage patterns.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted project data (source code, markdown files) to generate audits and critiques, which could be used to influence the agent's behavior.
Ingestion points: Reads project files and a local PRODUCT.md context file.
Boundary markers: None; the skill does not use specific delimiters to separate user data from instructions during analysis.
Capability inventory: Subprocess execution (native binary), shell command execution (npx), and file write operations (DESIGN.md).
Sanitization: No evidence of content sanitization or instruction filtering for ingested project data.

impeccable