feat: add mm-skill for multi-modal directory indexing and exploration#10
feat: add mm-skill for multi-modal directory indexing and exploration#10
Conversation
Co-Authored-By: Sudeep Pillai <sudeep.pillai@gmail.com>
🤖 Devin AI EngineerI'll be helping with this pull request! Here's what you should know: ✅ I will automatically:
Note: I can only respond to comments from users who have write access to this repository. ⚙️ Control Options:
|
![devin-ai-integration[bot]](https://avatars.githubusercontent.com/in/811515?s=60&v=4){kind=small}
![gemini-code-assist[bot]](https://avatars.githubusercontent.com/in/956858?s=60&v=4){kind=small}
There was a problem hiding this comment.
Code Review
This pull request introduces the mm-skill, a multi-modal context management CLI, by adding it to the marketplace configuration and providing a detailed SKILL.md guide. The documentation covers various commands for indexing, searching, and extracting content from diverse file types. Feedback focuses on correcting a contradiction regarding document extraction processing levels and consolidating several redundant entries in the tips and modes sections of the documentation.
| Level 1 behavior by file type (<100ms target): | ||
|
|
||
| - **PDF** (.pdf): text extraction via pypdfium2. Scanned/image-only PDFs return empty. | ||
| - **Document** (.docx, .pptx): markdown extraction via docling at L2. |
There was a problem hiding this comment.
This item is listed under Level 1 behavior but explicitly mentions at L2. Additionally, docling extraction typically takes longer than the 100ms target mentioned in the section header. Please clarify the level and performance expectations for this file type.
| - Start with `find --tree --depth 1` then `wc --by-kind` for the fastest directory overview. | ||
| - Use `mm find . --tree` to explore project structure. | ||
| - Use `--format json` when you need to parse output programmatically. | ||
| - `find` returns paths only when piped, else it returns full metadata rows. | ||
| - For exhaustive path lists, prefer `mm find <dir> --columns path | tail -n +2` and pipe/redirect as needed. | ||
| - Start with `mm find <dir> --tree --depth 1` then `mm wc <dir> --by-kind` for the fastest directory overview. | ||
| - Use `--format json` when you need to parse output programmatically. | ||
| - `find` returns paths only when piped, full metadata rows in TTY. |
There was a problem hiding this comment.
This section contains several redundant tips. Lines 288-290 repeat information already provided in lines 283-286. Consolidating them improves the clarity of the instructions for the agent.
| - Start with `find --tree --depth 1` then `wc --by-kind` for the fastest directory overview. | |
| - Use `mm find . --tree` to explore project structure. | |
| - Use `--format json` when you need to parse output programmatically. | |
| - `find` returns paths only when piped, else it returns full metadata rows. | |
| - For exhaustive path lists, prefer `mm find <dir> --columns path | tail -n +2` and pipe/redirect as needed. | |
| - Start with `mm find <dir> --tree --depth 1` then `mm wc <dir> --by-kind` for the fastest directory overview. | |
| - Use `--format json` when you need to parse output programmatically. | |
| - `find` returns paths only when piped, full metadata rows in TTY. | |
| - Start with mm find <dir> --tree --depth 1 then mm wc <dir> --by-kind for the fastest directory overview. | |
| - Use mm find . --tree to explore project structure. | |
| - Use --format json when you need to parse output programmatically. | |
| - find returns paths only when piped, full metadata rows in TTY. | |
| - For exhaustive path lists, prefer mm find <dir> --columns path | tail -n +2 and pipe/redirect as needed. |
| - L2 modes: (default, simple LLM captions), `--mode fast` (10 words + tags), `--mode accurate` (200 words + tags). | ||
| - Use `--mode fast` for quick L2 summaries (10 words), `--mode accurate` for detailed ones (200 words). |
There was a problem hiding this comment.
These two tips provide redundant information about the L2 modes. They should be merged into a single concise instruction.
| - L2 modes: (default, simple LLM captions), `--mode fast` (10 words + tags), `--mode accurate` (200 words + tags). | |
| - Use `--mode fast` for quick L2 summaries (10 words), `--mode accurate` for detailed ones (200 words). | |
| - L2 modes: (default, simple LLM captions), --mode fast (10 words + tags), --mode accurate (200 words + tags). |
1 participant
Summary
Adds a new
mm-skillto the skills repository. This skill provides agent instructions for themmCLI — a high-performance multi-modal context management tool for indexing, exploring, querying, and extracting content from directories containing images, videos, PDFs, code, and other files.Changes:
skills/mm-skill/SKILL.mdwith full command reference (find,cat,grep,sql,wc,bench,config,profile).claude-plugin/marketplace.jsonREADME.mdReview & Testing Checklist for Human
skills/mm-skill/SKILL.mdcontent to confirm it is appropriate for this public repository and does not contain any sensitive internal detailshttps://vlm-run.github.io/mm/install/install.sh) is live and correctmarketplace.jsonparses as valid JSON and the new plugin entry looks correctNotes
Link to Devin session: https://app.devin.ai/sessions/b476c7817877422383b2f73aa1381bd6
Requested by: @spillai