The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill uses recursive grep commands in Step 0 to search the user's ~/.dotfiles/ directory for hardcoded secrets, including API keys, tokens, secrets, passwords, and private key headers. The captured results, including the file paths and content of detected secrets, are ingested into the agent's context. It also specifically audits the file permissions of identity configuration files to ensure they are properly restricted.
[COMMAND_EXECUTION]: The skill executes a wide range of shell utilities to collect system state and performance metrics. Commands including stat for file permissions, time zsh for startup benchmarks, find for symlink integrity, nvim for plugin load profiling, and which for tool inventory are executed across the audit steps.
[DYNAMIC_EXECUTION]: In Step 1, the skill performs dynamic execution of local code by iterating through all shell scripts located in ~/.zsh.d/ and sourcing each file within a subshell to profile its individual impact on startup time. This action involves executing user-controlled code found on the local filesystem.
[INDIRECT_PROMPT_INJECTION]: The skill processes raw data from the filesystem and command outputs, which could potentially contain malicious instructions designed to influence the agent's subsequent logic or report generation.
Ingestion points: content identified by grep in ~/.dotfiles/, broken link paths from find, tool location strings from which, and Neovim performance logs from /tmp/nvim-startup.log.
Boundary markers: None are present; filesystem data is ingested and processed without delimiters or explicit instruction to ignore embedded commands.
Capability inventory: Extensive local command execution and file reading capabilities.
Sanitization: No sanitization, validation, or escaping of the ingested filesystem content is performed.

audit