Skip to content

add samkshipta (skill) and claude-skill-linter (tool)#591

Open
MukundaKatta wants to merge 1 commit intotravisvn:mainfrom
MukundaKatta:add-samkshipta-and-linter
Open

add samkshipta (skill) and claude-skill-linter (tool)#591
MukundaKatta wants to merge 1 commit intotravisvn:mainfrom
MukundaKatta:add-samkshipta-and-linter

Conversation

@MukundaKatta
Copy link
Copy Markdown

What

Two additions:

1. Individual Skills → samkshipta

Repo: https://github.com/MukundaKatta/samkshipta (v0.1.0, MIT)

Sanskrit-inspired token-compression skill. Three intensity levels:

Level Savings Use when
samkshipta-lite ~30% Mixed audience, teaching
samkshipta ~60-70% Expert peer, code review
samkshipta-sutra ~75-85% Commit messages, expert-only

Formal register throughout — the boardroom sibling of JuliusBrussee/caveman's meme-informal take on the same compression idea. Caveman credited as inspiration in the README.

Includes two slash commands (/samkshipta-commit, /samkshipta-review), bash installer, and a documented benchmarking methodology.

2. Tools → claude-skill-linter

Repo: https://github.com/MukundaKatta/claude-skill-linter (v0.1.0, MIT)

GitHub Action that lints SKILL.md files. Catches:

  • ❌ Missing required frontmatter (name, description)
  • ❌ Malformed/unparseable YAML
  • ❌ Hardcoded secrets (sk-, ghp_, AIza, xoxb-, github_pat_)
  • ❌ Non-kebab-case name
  • ⚠️ Description lacking trigger language
  • ⚠️ Broken relative-path references
  • ⚠️ Unknown frontmatter fields (typo detection)

Deployable as a one-line uses: in any repo with SKILL.md files. Currently running in CI against MukundaKatta/samkshipta and submitted as a PR (#172) to add it to JuliusBrussee/caveman.

Disclosure

I'm the author of both. Additions follow the non-promotional guideline — both are genuinely free (MIT), standalone, and solve real problems (token cost for saṃkṣipta; malformed SKILL.md for the linter). Happy to remove either if maintainers don't think they fit.

Test plan

  • Links resolve and point to public MIT repos
  • saṃkṣipta passes its own lint in strict mode (dogfooded via CI)
  • claude-skill-linter has CI workflow + fixture tests (3/3 passing)
  • Formatting matches existing table/list style in the list

samkshipta — https://github.com/MukundaKatta/samkshipta
  Sanskrit-inspired token-compression skill. Three intensity levels
  (lite / samkshipta / sutra). Formal register — suits code review
  and client-facing contexts where caveman-speak is inappropriate.
  v0.1.0 released, MIT, all 3 SKILL.md files pass claude-skill-linter
  in strict mode.

claude-skill-linter — https://github.com/MukundaKatta/claude-skill-linter
  GitHub Action that validates SKILL.md frontmatter, detects hardcoded
  API-key patterns, and flags missing required fields. Both self-test
  CI and dogfooded on samkshipta. v0.1.0 released, MIT.

Signed-off-by: Mukunda Katta <mukunda.vjcs6@gmail.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant