caveman-compress
Compress natural language memory files (CLAUDE.md, todos, preferences) into caveman format to save input tokens. Preserves all technical substance, code, URLs, and structure. Compressed version overwrites the original file. Human-readable backup saved as FILE.original.md. Trigger: /caveman-compress FILEPATH or "compress memory file"
72
88%
Does it follow best practices?
Impact
—
No eval scenarios have been run
Risky
Do not use without reviewing
Quality
Discovery
100%Based on the skill's description, can an agent find and select it at the right time? Clear, specific descriptions lead to better discovery.
This is an excellent skill description that clearly communicates what the skill does (compresses memory files into caveman format to save tokens), how it works (overwrites original, saves backup), and when to use it (explicit trigger command and natural language phrase). It uses third person voice throughout and provides specific, concrete details without unnecessary verbosity.
| Dimension | Reasoning | Score |
|---|---|---|
Specificity | Lists multiple specific concrete actions: compress natural language memory files, preserve technical substance/code/URLs/structure, overwrite original file, save human-readable backup as FILE.original.md. Very detailed about what happens. | 3 / 3 |
Completeness | Clearly answers both 'what' (compress natural language memory files into caveman format, preserving technical substance, creating backups) and 'when' (explicit trigger clause: '/caveman-compress FILEPATH or "compress memory file"'). | 3 / 3 |
Trigger Term Quality | Includes natural trigger terms users would say: '/caveman-compress', 'compress memory file', 'CLAUDE.md', 'todos', 'preferences', 'memory files', 'input tokens'. Good coverage of both command-style and natural language triggers. | 3 / 3 |
Distinctiveness Conflict Risk | Very distinct niche — 'caveman format' compression of memory files is highly specific and unlikely to conflict with other skills. The explicit command trigger '/caveman-compress' and specific file types (CLAUDE.md, todos, preferences) make it clearly distinguishable. | 3 / 3 |
Total | 12 / 12 Passed |
Implementation
77%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
This is a well-structured, actionable skill with clear workflow steps, validation checkpoints, and error recovery. The compression rules are comprehensive and the before/after examples are helpful. The main weakness is moderate verbosity in the rules sections—some lists (filler words, hedging phrases) could be condensed since Claude already understands these linguistic categories.
Suggestions
Condense the Remove section by using category descriptions rather than exhaustive word lists (e.g., 'Articles, filler adverbs, pleasantries, hedging phrases, redundant connectives') since Claude already knows what these categories contain.
Consider merging the two compression examples into one, as they demonstrate the same pattern (removing filler, shortening prose).
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The compression rules section is thorough but somewhat verbose for Claude—listing every filler word and hedging phrase is arguably unnecessary since Claude already understands these categories. The boundaries and preserve sections are well-organized but could be tighter. The examples are useful but two showing the same pattern is slightly redundant. | 2 / 3 |
Actionability | The skill provides a concrete CLI command (`python3 -m scripts <absolute_filepath>`), clear before/after compression examples, explicit lists of what to remove/preserve/compress, and specific file extension boundaries. Claude knows exactly what to do at each step. | 3 / 3 |
Workflow Clarity | The process section has a clear numbered sequence with explicit validation, error recovery (cherry-pick fix, retry up to 2 times), and a failure fallback (leave original untouched). The feedback loop for validation errors is well-defined. | 3 / 3 |
Progressive Disclosure | The content is well-structured with clear headings and sections, but the compression rules are quite detailed inline. The rules for preserve/remove/compress could potentially be split into a reference file. However, for a skill of this size (~100 lines), keeping it in one file is reasonable. References to `scripts/__main__.py` are present but no bundle files were provided to verify. | 2 / 3 |
Total | 10 / 12 Passed |
Validation
100%Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.
Validation — 11 / 11 Passed
Validation for skill structure
No warnings or errors.
- Repository
- JuliusBrussee/caveman
- Commit
655b7d9
- Reviewed
Table of Contents
Is this your skill?
If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.