skill-condenser

Name/description match the SKILL.md instructions: the skill reads a SKILL.md, runs iterative 'cod-iteration' passes to compress content, and outputs a condensed SKILL.md. There are no requested binaries, env vars, or config paths that would be unrelated to this purpose.

Instructions explicitly tell the agent to read and rewrite SKILL.md content (including moving detailed sections to references/). That's expected for a condenser, but the SKILL.md is somewhat vague about where the source file comes from and how writes are authorized. It also references an undefined step/tool 'cod-iteration' (presumably a model/internal routine). The skill does not instruct sending data to external endpoints, but you should confirm how 'cod-iteration' is implemented so you know whether content is kept locally or sent externally. Frontmatter preservation is required; frontmatter can occasionally contain sensitive metadata, so the agent's handling of it should be confirmed.

No install spec and no code files — instruction-only. This is low risk because nothing is written to disk by an installer. The runtime behavior depends on the agent's own capabilities (no external downloads or package installs are requested).

The skill requests no environment variables or credentials, which is proportional. A minor caution: frontmatter or example SKILL.md files could embed tokens or URLs; the instructions preserve frontmatter exactly rather than redacting it, so verify that frontmatter does not contain secrets before running automated condensation that may write those fields back or expose them in logs.

always is false and there is no claim of modifying other skills' configurations or system-wide agent settings. The instructions do describe modifying repository files (moving sections to references/ and updating SKILL.md), which is within scope for a documentation refactorer; just ensure you consent to those repository writes.