CHM Reader Skill

Read CHM (compiled HTML Help — .chm) files: software help files, offline manuals. Two modes, mirroring tool-pdf-reader: chunked stdout (read topics as you go, costs context) and full-dump to a markdown file (token-cheap, re-readable via Read/Grep).

Zero pip dependencies. Decompiles with 7-Zip (7z, checked on PATH + C:\Program Files\7-Zip) or falls back to Windows' built-in hh.exe -decompile. Extraction is cached in %TEMP%/chm_tool/ keyed by file identity, so repeated commands don't re-decompile.

SKILL_DIR below = this skill's directory (skills/tool-chm-reader/ in the repo, or ~/.claude/skills/tool-chm-reader/).

Usage

1. Get CHM info

python "<SKILL_DIR>/chm_tool.py" info "path/to/file.chm"

Returns: size, title, HTML page count, TOC entry count, cache dir.

2. Table of contents

python "<SKILL_DIR>/chm_tool.py" toc "path/to/file.chm"

Numbered, indentation shows hierarchy. The numbers are what extract --start/--end addresses. CHMs without a .hhc sitemap fall back to a flat page list (same numbering).

3. Extract topics (TOC-number range, stdout)

python "<SKILL_DIR>/chm_tool.py" extract "file.chm" --start 11 --end 30

Read 10–30 topics at a time. Use when triaging which sections matter, or you only need one chapter.

4. One page by internal path

python "<SKILL_DIR>/chm_tool.py" page "file.chm" bonding.htm

5. Search

python "<SKILL_DIR>/chm_tool.py" search "file.chm" "cross bonding"

Case-insensitive, whole archive; prints TOC entry + matching line snippets. Search is literal — try both crossbonding and cross bonding style variants for compound terms.

6. Dump the whole CHM to markdown (token-saver)

python "<SKILL_DIR>/chm_tool.py" dump "file.chm" "path/to/output.md"

One markdown file, TOC-ordered, heading depth mirroring the TOC hierarchy, <!-- source: page.htm --> markers per topic, non-TOC pages appended under "Unlisted pages".

• Use when the CHM will be referenced more than once: dump once, then Read/Grep the .md in this and future sessions.
• For a vault CHM, dump beside it: Foo/bar.chm → Foo/bar-extracted.md (matches the *-extracted convention from tool-pdf-reader).
• A 381-topic / 18 MB CHM dumps to ~490 KB of markdown in seconds.

Workflow for digesting a software help file

1. info — size + page count.
2. toc — understand the manual's structure; pick the chapters that matter.
3. dump to a sibling *-extracted.md — the durable artifact.
4. Read/Grep the .md; hand-write digests or KB notes from it (KB digestion itself belongs to the KB skills, not this one).

Notes

• Content conversion is stdlib HTML→markdown (headings, lists, tables, pre, bold/italic, image alt text). Screenshots/toolbar-icon images render as nothing or ![alt] — CHM manuals lean on inline images, so occasional "click on ." gaps are the image placeholders, not extraction bugs.
• CHM pages are commonly windows-1252; the reader honours each page's meta charset.
• .chm files downloaded from the internet may need "Unblock" in file Properties before Windows' viewer renders them — this tool reads them regardless (7z/hh don't care about the zone marker).
• Fresh Linux/mac clones need 7z (p7zip) installed; Windows works out of the box via hh.exe even without 7-Zip.

Trigger phrases

User says	Action
"read this chm" / "what's in this help file"	info → toc → extract chunks
"search the help for X"	search
"dump the chm" / "save it as markdown"	dump to sibling *-extracted.md
hands over software-info zip with .chm inside	extract zip → dump each .chm