feat(skills): fix browse cap, add source links + copy buttons + category cleanup - #37143
Merged
Conversation
…ory cleanup Skills discovery surfaced ~136 of 88k skills in the CLI and gave community skills no clickable source on the docs page. Three coupled fixes: CLI browse: - hermes skills browse capped at 50 because the per-source limit dict had no 'hermes-index' key — when the centralized index is available the router skips external APIs and serves only the index, so the default-50 fallthrough silently truncated the whole hub. Add hermes-index: 5000. Browse now loads 5367 (269 pages) instead of 136. - Add an Identifier column + install/inspect hint to the browse table so users can act on what they see without a second 'search'. - Route the TUI browse_skills() helper through parallel_search_sources so it inherits the same index-aware source-skip (was double-counting); expose identifier in its output. Docs Skills Hub page: - Synthesize a sourceUrl for every community skill (github tree URL, clawhub / skills.sh / lobehub / browse.sh detail pages), preferring the adapter's explicit extra.detail_url/source_url/repo_url. Expanded cards now show 'View source' for community skills (was nothing) and keep 'View full documentation' for built-in/optional. 99% coverage. - Add a Copy button on the install command. - Add a loading state instead of flashing '0 skills / No skills found' while the 45MB catalog fetches. Category cleanup: - _guess_category fell back to tags[0] verbatim, producing ~430 junk one-off categories (version strings, brand names: '0.10.7 Dev', 'Doramagic Crystal'). Now only curated buckets are accepted; unknowns fold into 'Other'. Widen the tag->category map so common community tags route to real buckets. 430 -> 173 categories, top 20 all meaningful. Tests: tests/website/test_extract_skills.py covers _source_url synthesis + precedence and _guess_category curation (13 tests). All 27 skills-hub CLI tests still pass. Docusaurus build verified; expanded cards confirmed in browser for both community (View source) and built-in (View full docs).
Contributor
🔎 Lint report:
|
| Rule | Count |
|---|---|
unresolved-import |
1 |
First entries
tests/website/test_extract_skills.py:22: [unresolved-import] unresolved-import: Cannot resolve imported module `pytest`
✅ Fixed issues: none
Unchanged: 4977 pre-existing issues carried over.
Diagnostics are surfaced as warnings — this check never fails the build.
1 task
changman
pushed a commit
to changman/hermes-agent
that referenced
this pull request
Jun 10, 2026
…ory cleanup (NousResearch#37143) Skills discovery surfaced ~136 of 88k skills in the CLI and gave community skills no clickable source on the docs page. Three coupled fixes: CLI browse: - hermes skills browse capped at 50 because the per-source limit dict had no 'hermes-index' key — when the centralized index is available the router skips external APIs and serves only the index, so the default-50 fallthrough silently truncated the whole hub. Add hermes-index: 5000. Browse now loads 5367 (269 pages) instead of 136. - Add an Identifier column + install/inspect hint to the browse table so users can act on what they see without a second 'search'. - Route the TUI browse_skills() helper through parallel_search_sources so it inherits the same index-aware source-skip (was double-counting); expose identifier in its output. Docs Skills Hub page: - Synthesize a sourceUrl for every community skill (github tree URL, clawhub / skills.sh / lobehub / browse.sh detail pages), preferring the adapter's explicit extra.detail_url/source_url/repo_url. Expanded cards now show 'View source' for community skills (was nothing) and keep 'View full documentation' for built-in/optional. 99% coverage. - Add a Copy button on the install command. - Add a loading state instead of flashing '0 skills / No skills found' while the 45MB catalog fetches. Category cleanup: - _guess_category fell back to tags[0] verbatim, producing ~430 junk one-off categories (version strings, brand names: '0.10.7 Dev', 'Doramagic Crystal'). Now only curated buckets are accepted; unknowns fold into 'Other'. Widen the tag->category map so common community tags route to real buckets. 430 -> 173 categories, top 20 all meaningful. Tests: tests/website/test_extract_skills.py covers _source_url synthesis + precedence and _guess_category curation (13 tests). All 27 skills-hub CLI tests still pass. Docusaurus build verified; expanded cards confirmed in browser for both community (View source) and built-in (View full docs).
alt-glitch
pushed a commit
that referenced
this pull request
Jun 14, 2026
…ory cleanup (#37143) Skills discovery surfaced ~136 of 88k skills in the CLI and gave community skills no clickable source on the docs page. Three coupled fixes: CLI browse: - hermes skills browse capped at 50 because the per-source limit dict had no 'hermes-index' key — when the centralized index is available the router skips external APIs and serves only the index, so the default-50 fallthrough silently truncated the whole hub. Add hermes-index: 5000. Browse now loads 5367 (269 pages) instead of 136. - Add an Identifier column + install/inspect hint to the browse table so users can act on what they see without a second 'search'. - Route the TUI browse_skills() helper through parallel_search_sources so it inherits the same index-aware source-skip (was double-counting); expose identifier in its output. Docs Skills Hub page: - Synthesize a sourceUrl for every community skill (github tree URL, clawhub / skills.sh / lobehub / browse.sh detail pages), preferring the adapter's explicit extra.detail_url/source_url/repo_url. Expanded cards now show 'View source' for community skills (was nothing) and keep 'View full documentation' for built-in/optional. 99% coverage. - Add a Copy button on the install command. - Add a loading state instead of flashing '0 skills / No skills found' while the 45MB catalog fetches. Category cleanup: - _guess_category fell back to tags[0] verbatim, producing ~430 junk one-off categories (version strings, brand names: '0.10.7 Dev', 'Doramagic Crystal'). Now only curated buckets are accepted; unknowns fold into 'Other'. Widen the tag->category map so common community tags route to real buckets. 430 -> 173 categories, top 20 all meaningful. Tests: tests/website/test_extract_skills.py covers _source_url synthesis + precedence and _guess_category curation (13 tests). All 27 skills-hub CLI tests still pass. Docusaurus build verified; expanded cards confirmed in browser for both community (View source) and built-in (View full docs).
kossteg
pushed a commit
to kossteg/hermes-agent
that referenced
this pull request
Jun 16, 2026
…ory cleanup (NousResearch#37143) Skills discovery surfaced ~136 of 88k skills in the CLI and gave community skills no clickable source on the docs page. Three coupled fixes: CLI browse: - hermes skills browse capped at 50 because the per-source limit dict had no 'hermes-index' key — when the centralized index is available the router skips external APIs and serves only the index, so the default-50 fallthrough silently truncated the whole hub. Add hermes-index: 5000. Browse now loads 5367 (269 pages) instead of 136. - Add an Identifier column + install/inspect hint to the browse table so users can act on what they see without a second 'search'. - Route the TUI browse_skills() helper through parallel_search_sources so it inherits the same index-aware source-skip (was double-counting); expose identifier in its output. Docs Skills Hub page: - Synthesize a sourceUrl for every community skill (github tree URL, clawhub / skills.sh / lobehub / browse.sh detail pages), preferring the adapter's explicit extra.detail_url/source_url/repo_url. Expanded cards now show 'View source' for community skills (was nothing) and keep 'View full documentation' for built-in/optional. 99% coverage. - Add a Copy button on the install command. - Add a loading state instead of flashing '0 skills / No skills found' while the 45MB catalog fetches. Category cleanup: - _guess_category fell back to tags[0] verbatim, producing ~430 junk one-off categories (version strings, brand names: '0.10.7 Dev', 'Doramagic Crystal'). Now only curated buckets are accepted; unknowns fold into 'Other'. Widen the tag->category map so common community tags route to real buckets. 430 -> 173 categories, top 20 all meaningful. Tests: tests/website/test_extract_skills.py covers _source_url synthesis + precedence and _guess_category curation (13 tests). All 27 skills-hub CLI tests still pass. Docusaurus build verified; expanded cards confirmed in browser for both community (View source) and built-in (View full docs).
xyshanren
pushed a commit
to xyshanren/hermes-agent-cn
that referenced
this pull request
Jun 17, 2026
…ory cleanup (NousResearch#37143) Skills discovery surfaced ~136 of 88k skills in the CLI and gave community skills no clickable source on the docs page. Three coupled fixes: CLI browse: - hermes skills browse capped at 50 because the per-source limit dict had no 'hermes-index' key — when the centralized index is available the router skips external APIs and serves only the index, so the default-50 fallthrough silently truncated the whole hub. Add hermes-index: 5000. Browse now loads 5367 (269 pages) instead of 136. - Add an Identifier column + install/inspect hint to the browse table so users can act on what they see without a second 'search'. - Route the TUI browse_skills() helper through parallel_search_sources so it inherits the same index-aware source-skip (was double-counting); expose identifier in its output. Docs Skills Hub page: - Synthesize a sourceUrl for every community skill (github tree URL, clawhub / skills.sh / lobehub / browse.sh detail pages), preferring the adapter's explicit extra.detail_url/source_url/repo_url. Expanded cards now show 'View source' for community skills (was nothing) and keep 'View full documentation' for built-in/optional. 99% coverage. - Add a Copy button on the install command. - Add a loading state instead of flashing '0 skills / No skills found' while the 45MB catalog fetches. Category cleanup: - _guess_category fell back to tags[0] verbatim, producing ~430 junk one-off categories (version strings, brand names: '0.10.7 Dev', 'Doramagic Crystal'). Now only curated buckets are accepted; unknowns fold into 'Other'. Widen the tag->category map so common community tags route to real buckets. 430 -> 173 categories, top 20 all meaningful. Tests: tests/website/test_extract_skills.py covers _source_url synthesis + precedence and _guess_category curation (13 tests). All 27 skills-hub CLI tests still pass. Docusaurus build verified; expanded cards confirmed in browser for both community (View source) and built-in (View full docs). (cherry picked from commit 59510d7)
T02200059
pushed a commit
to T02200059/hermes-agent
that referenced
this pull request
Jun 18, 2026
…ory cleanup (NousResearch#37143) Skills discovery surfaced ~136 of 88k skills in the CLI and gave community skills no clickable source on the docs page. Three coupled fixes: CLI browse: - hermes skills browse capped at 50 because the per-source limit dict had no 'hermes-index' key — when the centralized index is available the router skips external APIs and serves only the index, so the default-50 fallthrough silently truncated the whole hub. Add hermes-index: 5000. Browse now loads 5367 (269 pages) instead of 136. - Add an Identifier column + install/inspect hint to the browse table so users can act on what they see without a second 'search'. - Route the TUI browse_skills() helper through parallel_search_sources so it inherits the same index-aware source-skip (was double-counting); expose identifier in its output. Docs Skills Hub page: - Synthesize a sourceUrl for every community skill (github tree URL, clawhub / skills.sh / lobehub / browse.sh detail pages), preferring the adapter's explicit extra.detail_url/source_url/repo_url. Expanded cards now show 'View source' for community skills (was nothing) and keep 'View full documentation' for built-in/optional. 99% coverage. - Add a Copy button on the install command. - Add a loading state instead of flashing '0 skills / No skills found' while the 45MB catalog fetches. Category cleanup: - _guess_category fell back to tags[0] verbatim, producing ~430 junk one-off categories (version strings, brand names: '0.10.7 Dev', 'Doramagic Crystal'). Now only curated buckets are accepted; unknowns fold into 'Other'. Widen the tag->category map so common community tags route to real buckets. 430 -> 173 categories, top 20 all meaningful. Tests: tests/website/test_extract_skills.py covers _source_url synthesis + precedence and _guess_category curation (13 tests). All 27 skills-hub CLI tests still pass. Docusaurus build verified; expanded cards confirmed in browser for both community (View source) and built-in (View full docs).
waefrebeorn
pushed a commit
to waefrebeorn/slermes
that referenced
this pull request
Jul 2, 2026
…ory cleanup (NousResearch#37143) Skills discovery surfaced ~136 of 88k skills in the CLI and gave community skills no clickable source on the docs page. Three coupled fixes: CLI browse: - hermes skills browse capped at 50 because the per-source limit dict had no 'hermes-index' key — when the centralized index is available the router skips external APIs and serves only the index, so the default-50 fallthrough silently truncated the whole hub. Add hermes-index: 5000. Browse now loads 5367 (269 pages) instead of 136. - Add an Identifier column + install/inspect hint to the browse table so users can act on what they see without a second 'search'. - Route the TUI browse_skills() helper through parallel_search_sources so it inherits the same index-aware source-skip (was double-counting); expose identifier in its output. Docs Skills Hub page: - Synthesize a sourceUrl for every community skill (github tree URL, clawhub / skills.sh / lobehub / browse.sh detail pages), preferring the adapter's explicit extra.detail_url/source_url/repo_url. Expanded cards now show 'View source' for community skills (was nothing) and keep 'View full documentation' for built-in/optional. 99% coverage. - Add a Copy button on the install command. - Add a loading state instead of flashing '0 skills / No skills found' while the 45MB catalog fetches. Category cleanup: - _guess_category fell back to tags[0] verbatim, producing ~430 junk one-off categories (version strings, brand names: '0.10.7 Dev', 'Doramagic Crystal'). Now only curated buckets are accepted; unknowns fold into 'Other'. Widen the tag->category map so common community tags route to real buckets. 430 -> 173 categories, top 20 all meaningful. Tests: tests/website/test_extract_skills.py covers _source_url synthesis + precedence and _guess_category curation (13 tests). All 27 skills-hub CLI tests still pass. Docusaurus build verified; expanded cards confirmed in browser for both community (View source) and built-in (View full docs).
santhreal
pushed a commit
to santhreal/hermes-agent
that referenced
this pull request
Jul 13, 2026
…ory cleanup (NousResearch#37143) Skills discovery surfaced ~136 of 88k skills in the CLI and gave community skills no clickable source on the docs page. Three coupled fixes: CLI browse: - hermes skills browse capped at 50 because the per-source limit dict had no 'hermes-index' key — when the centralized index is available the router skips external APIs and serves only the index, so the default-50 fallthrough silently truncated the whole hub. Add hermes-index: 5000. Browse now loads 5367 (269 pages) instead of 136. - Add an Identifier column + install/inspect hint to the browse table so users can act on what they see without a second 'search'. - Route the TUI browse_skills() helper through parallel_search_sources so it inherits the same index-aware source-skip (was double-counting); expose identifier in its output. Docs Skills Hub page: - Synthesize a sourceUrl for every community skill (github tree URL, clawhub / skills.sh / lobehub / browse.sh detail pages), preferring the adapter's explicit extra.detail_url/source_url/repo_url. Expanded cards now show 'View source' for community skills (was nothing) and keep 'View full documentation' for built-in/optional. 99% coverage. - Add a Copy button on the install command. - Add a loading state instead of flashing '0 skills / No skills found' while the 45MB catalog fetches. Category cleanup: - _guess_category fell back to tags[0] verbatim, producing ~430 junk one-off categories (version strings, brand names: '0.10.7 Dev', 'Doramagic Crystal'). Now only curated buckets are accepted; unknowns fold into 'Other'. Widen the tag->category map so common community tags route to real buckets. 430 -> 173 categories, top 20 all meaningful. Tests: tests/website/test_extract_skills.py covers _source_url synthesis + precedence and _guess_category curation (13 tests). All 27 skills-hub CLI tests still pass. Docusaurus build verified; expanded cards confirmed in browser for both community (View source) and built-in (View full docs).
donbowman
pushed a commit
to donbowman/hermes-agent
that referenced
this pull request
Jul 13, 2026
…ory cleanup (NousResearch#37143) Skills discovery surfaced ~136 of 88k skills in the CLI and gave community skills no clickable source on the docs page. Three coupled fixes: CLI browse: - hermes skills browse capped at 50 because the per-source limit dict had no 'hermes-index' key — when the centralized index is available the router skips external APIs and serves only the index, so the default-50 fallthrough silently truncated the whole hub. Add hermes-index: 5000. Browse now loads 5367 (269 pages) instead of 136. - Add an Identifier column + install/inspect hint to the browse table so users can act on what they see without a second 'search'. - Route the TUI browse_skills() helper through parallel_search_sources so it inherits the same index-aware source-skip (was double-counting); expose identifier in its output. Docs Skills Hub page: - Synthesize a sourceUrl for every community skill (github tree URL, clawhub / skills.sh / lobehub / browse.sh detail pages), preferring the adapter's explicit extra.detail_url/source_url/repo_url. Expanded cards now show 'View source' for community skills (was nothing) and keep 'View full documentation' for built-in/optional. 99% coverage. - Add a Copy button on the install command. - Add a loading state instead of flashing '0 skills / No skills found' while the 45MB catalog fetches. Category cleanup: - _guess_category fell back to tags[0] verbatim, producing ~430 junk one-off categories (version strings, brand names: '0.10.7 Dev', 'Doramagic Crystal'). Now only curated buckets are accepted; unknowns fold into 'Other'. Widen the tag->category map so common community tags route to real buckets. 430 -> 173 categories, top 20 all meaningful. Tests: tests/website/test_extract_skills.py covers _source_url synthesis + precedence and _guess_category curation (13 tests). All 27 skills-hub CLI tests still pass. Docusaurus build verified; expanded cards confirmed in browser for both community (View source) and built-in (View full docs).
Gravezzz
pushed a commit
to Gravezzz/hermes-agent
that referenced
this pull request
Jul 21, 2026
…ory cleanup (NousResearch#37143) Skills discovery surfaced ~136 of 88k skills in the CLI and gave community skills no clickable source on the docs page. Three coupled fixes: CLI browse: - hermes skills browse capped at 50 because the per-source limit dict had no 'hermes-index' key — when the centralized index is available the router skips external APIs and serves only the index, so the default-50 fallthrough silently truncated the whole hub. Add hermes-index: 5000. Browse now loads 5367 (269 pages) instead of 136. - Add an Identifier column + install/inspect hint to the browse table so users can act on what they see without a second 'search'. - Route the TUI browse_skills() helper through parallel_search_sources so it inherits the same index-aware source-skip (was double-counting); expose identifier in its output. Docs Skills Hub page: - Synthesize a sourceUrl for every community skill (github tree URL, clawhub / skills.sh / lobehub / browse.sh detail pages), preferring the adapter's explicit extra.detail_url/source_url/repo_url. Expanded cards now show 'View source' for community skills (was nothing) and keep 'View full documentation' for built-in/optional. 99% coverage. - Add a Copy button on the install command. - Add a loading state instead of flashing '0 skills / No skills found' while the 45MB catalog fetches. Category cleanup: - _guess_category fell back to tags[0] verbatim, producing ~430 junk one-off categories (version strings, brand names: '0.10.7 Dev', 'Doramagic Crystal'). Now only curated buckets are accepted; unknowns fold into 'Other'. Widen the tag->category map so common community tags route to real buckets. 430 -> 173 categories, top 20 all meaningful. Tests: tests/website/test_extract_skills.py covers _source_url synthesis + precedence and _guess_category curation (13 tests). All 27 skills-hub CLI tests still pass. Docusaurus build verified; expanded cards confirmed in browser for both community (View source) and built-in (View full docs).
leewenjie
pushed a commit
to leewenjie/hermes-agent
that referenced
this pull request
Aug 7, 2026
…ory cleanup (NousResearch#37143) Skills discovery surfaced ~136 of 88k skills in the CLI and gave community skills no clickable source on the docs page. Three coupled fixes: CLI browse: - hermes skills browse capped at 50 because the per-source limit dict had no 'hermes-index' key — when the centralized index is available the router skips external APIs and serves only the index, so the default-50 fallthrough silently truncated the whole hub. Add hermes-index: 5000. Browse now loads 5367 (269 pages) instead of 136. - Add an Identifier column + install/inspect hint to the browse table so users can act on what they see without a second 'search'. - Route the TUI browse_skills() helper through parallel_search_sources so it inherits the same index-aware source-skip (was double-counting); expose identifier in its output. Docs Skills Hub page: - Synthesize a sourceUrl for every community skill (github tree URL, clawhub / skills.sh / lobehub / browse.sh detail pages), preferring the adapter's explicit extra.detail_url/source_url/repo_url. Expanded cards now show 'View source' for community skills (was nothing) and keep 'View full documentation' for built-in/optional. 99% coverage. - Add a Copy button on the install command. - Add a loading state instead of flashing '0 skills / No skills found' while the 45MB catalog fetches. Category cleanup: - _guess_category fell back to tags[0] verbatim, producing ~430 junk one-off categories (version strings, brand names: '0.10.7 Dev', 'Doramagic Crystal'). Now only curated buckets are accepted; unknowns fold into 'Other'. Widen the tag->category map so common community tags route to real buckets. 430 -> 173 categories, top 20 all meaningful. Tests: tests/website/test_extract_skills.py covers _source_url synthesis + precedence and _guess_category curation (13 tests). All 27 skills-hub CLI tests still pass. Docusaurus build verified; expanded cards confirmed in browser for both community (View source) and built-in (View full docs).
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Skills discovery now surfaces the whole catalog and every skill has a clickable source + one-click install. Before:
hermes skills browseshowed ~136 of 88k skills, and 88k community skills on the docs page expanded to a card with no link to their origin.Changes
hermes skills browsecapped at 50 — the per-source limit dict had nohermes-indexkey, so when the centralized index is available (router skips external APIs and serves only the index) it hit the default-50 fallthrough and silently truncated the whole hub. Addedhermes-index: 5000→ browse loads 5,367 (269 pages) instead of 136.search.browse_skills()rerouted throughparallel_search_sources(was rolling its own loop → double-counted index + external sources); now inherits the index-aware skip and exposesidentifier.extract-skills.pysynthesizes asourceUrlfor every community skill (GitHub tree URL; clawhub / skills.sh / lobehub / browse.sh detail pages), preferring the adapter's explicitextra.detail_url/source_url/repo_url. Expanded cards show View source for community skills (was nothing), keep View full documentation for built-in/optional. 99% coverage._guess_categoryfell back totags[0]verbatim, producing ~430 junk one-off categories (0.10.7 Dev,Doramagic Crystal,Ap2). Now only curated buckets are accepted; unknowns fold into "Other". Widened the tag→category map. 430 → 173 categories, top 20 all meaningful.Validation
hermes skills browseskills loadedtests/website/test_extract_skills.py— 13 new tests (sourceUrl synthesis/precedence, category curation, junk-tag rejection).tests/hermes_cli/test_skills_hub.pypass.npx docusaurus buildsucceeds; expanded cards verified in-browser for both community (View source → clawhub.ai) and built-in (View full docs).Q&A addressed (from the audit request)
identifier(clawhub/x,skills-sh/owner/repo/x);_resolve_short_namedisambiguates with a Source/Trust/Identifier table. No change needed.Infographic