agentleFS
tyler-berggrentyler-berggren/tb-claude-kit/.claude/skills/parse/SKILL.md

parse

Skill2 starsChanged 5 days agoMITView on GitHub

Convert local documents (docx, pdf, xlsx, pptx, rtf, odt, epub, csv, html) to markdown via Firecrawl parse. Takes a file, a glob, or a directory and writes .md alongside the originals.

  • Reads credentials
  • Sends data out

1 other project uses a version of this.

Adopted by

Most recently taken up by

  • Other versions1 project uses a version of this.

    Word for word · 0

    None found.

    Edited · 0

    None found.

    Wrote their own version · 1

    1. enkerewpo/zoresearch★ 1same name

    How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.

  • In this repository24 other files it gives its agents.

The file

We split this skill into the rules it gives an agent

8 rules. About 0 are common, 5 are distinctive, and 3 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈5

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 3

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    If the path is missing or matches nothing, list the supported extensions and stop.

    line 29
  2. Distinctive:

    If none has one, say so and point at https://firecrawl.dev — do not guess at a key or fall back to keyless mode, which does not cover parse.

    line 45
  3. Distinctive:

    Prefer the MCP tool; fall back to curl.

    line 50
  4. Distinctive:

    2. Fallback — curl: Use this when the tool errors, when the MCP server is unavailable, or when parsing several files (one request each, run …

    line 62
  5. Very specific:

    Default: <original-basename>.md next to the source, originals untouched.

    line 79
  6. Distinctive:

    Do not claim the parse is clean without having looked at the output.

    line 92
  7. Very specific:

    Default timeout 30s, max 300s.

    line 97
  8. Very specific:

    Read .claude/kit.json → rules.parse if present and apply it.

    line 107

≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Parse — local documents to markdown

Firecrawl's /v2/parse endpoint turns a local binary document into clean markdown. It is the counterpart to /scrape: scrape is for public URLs, parse is for files on this machine that no URL points at — signed contracts, private decks, exported spreadsheets, scanned PDFs.

Docs: https://docs.firecrawl.dev/features/parse

Input

$ARGUMENTS — a path, then optional flags.

TokenMeaning
first tokenFile, directory, or glob. A directory means every supported file directly inside it — not recursive unless the user asks.
--formatsComma-separated: markdown (default), html, rawHtml, links, images, summary, json.
--out <dir>Write outputs here instead of alongside the originals.
--stdoutPrint the markdown into the conversation, write nothing.
--redactSet redactPII: true — strips personal identifiers from returned content.
--pages <n>PDFs only: cap at pdfOptions.maxPages.

If the path is missing or matches nothing, list the supported extensions and stop.

Supported: .pdf .docx .doc .docm .odt .rtf .html .htm .xhtml .xlsx .xls .xlsm .xlsb .ods .pptx .ppt .pptm .odp .epub .csv

A public document URL is not this skill — use firecrawl_scrape instead.

Credentials

Resolve a Firecrawl API key from the first source that has one, and never print it — assign to a variable, reference the variable:

  1. $FIRECRAWL_API_KEY already in the environment
  2. .env in the project root (set -a; . ./.env; set +a)
  3. .mcp.json → mcpServers.firecrawl.env.FIRECRAWL_API_KEY
  4. ~/.claude.json → the same path, under this project or globally

If none has one, say so and point at https://firecrawl.dev — do not guess at a key or fall back to keyless mode, which does not cover parse.

Method

Prefer the MCP tool; fall back to curl.

1. mcp__firecrawl__firecrawl_parse with filePath and formats. This works only when the firecrawl MCP server has FIRECRAWL_API_URL set — the local-mode handler hard-requires it and otherwise throws requires FIRECRAWL_API_URL to be set to a self-hosted Firecrawl API instance. It never checks that the URL is self-hosted, so setting it to https://api.firecrawl.dev satisfies the guard and routes to the cloud API. The kit's .mcp.json template sets this. If the tool throws that error here, the project's .mcp.json predates it: add the variable, mention that MCP servers only pick it up on session restart, and use curl for now.

2. Fallback — curl. Use this when the tool errors, when the MCP server is unavailable, or when parsing several files (one request each, run in parallel).

curl -sS -X POST https://api.firecrawl.dev/v2/parse \
  -H "Authorization: Bearer $KEY" \
  -F "file=@<path>" \
  -F 'options={"formats":["markdown"]};type=application/json'

Response is {success, data:{markdown, ...}}. On success: false, report the error verbatim rather than retrying blind. Parse the JSON with python, not jq-into-shell-quoting — document markdown is full of characters that break shell interpolation. Write the output file from python too.

Output

Default: <original-basename>.md next to the source, originals untouched. Never overwrite an existing .md without saying so first — check, and if one exists, ask unless the user passed an explicit --out.

Report a table of file → output → size. Then spot-check the conversion and say what you found:

  • Tables are where docx and pdf conversion degrades — open any output with fee schedules, phase plans, or line items and confirm the columns survived.
  • Multi-column PDFs can interleave text across columns.
  • Headers, footers, and page numbers often land mid-document.
  • Signature blocks and form underscores usually survive as literal \_\_\_.

Do not claim the parse is clean without having looked at the output.

Constraints

  • 50 MB per file, one file per request, no batch endpoint.
  • Default timeout 30s, max 300s. Long PDFs: bound with --pages.
  • Parse consumes Firecrawl credits per file.
  • Parsing uploads the file to Firecrawl's API. For anything confidential — contracts, private or personal data, anything under NDA — say so plainly before the first request. If the user already named the files, that is authorization; note it and proceed rather than blocking. Offer --redact when the documents contain personal information.

Project configuration

Read .claude/kit.json → rules.parse if present and apply it. Use it for project-specific filing conventions — where parsed documents belong, whether they get committed, whether a parsed contract should also be logged to the brain.

Absent a rule: keep the .md beside its original so the pair travels together, and offer a brain entry for documents worth remembering rather than assuming one.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.