The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the Bash tool to manage temporary files and execute local helper scripts (scripts/preflight.sh, scripts/parse-ratio.sh, scripts/validate-essence.sh, scripts/count-tokens.sh). All variable interpolation into shell commands, such as the SLUG variable, follows safe practices including character-set restriction via tr to prevent command injection.
[DATA_EXFILTRATION]: No network access is requested in the frontmatter (network: []), and no network-capable tools (like curl or wget) are used in the provided scripts. Data remains localized to the .galmuri/tmp directory.
[REMOTE_CODE_EXECUTION]: The skill does not perform remote downloads or execute code from untrusted external sources. All referenced scripts are part of the local skill package.
[PROMPT_INJECTION]: The LLM instructions are strictly focused on text analysis and transformation (Method 1-3). No instructions were found that attempt to bypass AI safety guidelines or extract system prompts.

distill