trkbt10/indexion-skills
9 skills4.6k installs9 starsGitHub
Install
npx skills add https://github.com/trkbt10/indexion-skillsSkills in this repo
1Indexion ReadmeIndexion README is a workflow skill for teams and solo maintainers who treat documentation as structured data, not a one-off markdown file. It covers the construction side of the Indexion toolchain: initializing template scaffolding, generating package-level READMEs from doc comments, planning writing tasks, assembling a root README from `docs/` fragments and per-package outputs according to `doc.json`, and validating consistency with `plan drift`. Repos vary—config may live at the root or under `.indexion/readme/`, and some MoonBit projects symlink `README.md` to `README.mbt.md` for doctest modules—so the skill insists you inspect `git diff`, symlinks, and `.indexion.toml` before editing. Prism pairs it with Build/docs as canonical placement while noting Ship review and Operate iterate moments when docs must stay aligned with code. For evaluating doc quality rather than building it, the SKILL.md points to the separate indexion-documentation skill. Ideal when you outgrow a single awesome-copilot README pass and need repeatable assembly across many packages.515installs2Indexion Kgfindexion-kgf is an agent skill for solo builders and maintainers who ship with indexion and need to see exactly how a KGF language spec treats a file. It wraps the indexion kgf CLI so you can list and update installed specs, add new ones, and drill into tokenization-only, event-only, or edge-only views—or run a full inspect pass when analysis output looks wrong. Use it while adding or fixing language support, when someone asks how indexion parses a path, or when a grep pattern fails because token kinds do not match expectations. The workflow is deliberately read-only inspection: you point at a file, optionally pin a spec, and interpret structured output rather than guessing from black-box graph results. That makes it practical for indie agent-tooling work where one person owns both the spec and the indexer behavior.511installs3Indexion SegmentIndexion Segment is an agent skill that teaches your coding agent to split long text into meaningful chunks using the indexion CLI—critical for solo builders who run their own RAG without managed chunking SaaS. It documents when to pick window divergence, TF-IDF, punctuation, or hybrid modes, plus flags for segment size targets, adaptive thresholds, and output prefixes. Reach for it while building ingestion for embeddings, preparing docs for similarity search, or splitting reports into processable sections. The skill is narrowly focused on the segmentation step rather than full vector DB deployment, which keeps it fast to invoke mid-pipeline. It assumes you will pair it with embedding and retrieval skills downstream in the same Build phase.511installs4Indexion Documentationindexion-documentation is an agent skill that drives the indexion CLI to answer what needs documentation and whether docs still match the code. Solo builders maintaining libraries or monorepos use it when public API surface grows faster than comments and READMEs. It reports overall coverage percentage with functions vs types splits, per-package inventory, and optional `--style=coverage` quick views. Full plans prioritize fixes; GitHub-issue and JSON formats support issue trackers and CI. Undocumented-item grep gives file-level targets without a full plan pass. The skill positions itself on the evaluation side of the documentation lifecycle—assessment and drift—while readme authoring stays in a sibling skill. Works wherever you have a checkout and indexion installed, especially before releases, onboarding contributors, or after large refactors when stale docs mislead agents and humans alike.509installs5Indexion Refactorindexion-refactor is an agent skill for solo builders who already ship features but feel structural debt accumulating—duplicate helpers, trivial wrappers, and the same domain rule reimplemented in three packages. After writing code, the skill walks you through indexion-driven detection at three levels: textual blocks via plan refactor, structural patterns via plan solid and plan unwrap, and conceptual duplication via explore plus human judgment. It is meant for post-change cleanup when a fix touched many files, when filesystem errors hint at path logic copied everywhere, or when you added format or I/O boundaries that should centralize. You detect with indexion, apply targeted extractions and deletions, then verify SoT holds. It complements feature skills rather than replacing design upfront; use it when merge-ready code still smells like copy-paste or compensating guards.509installs6Indexion Sddindexion-sdd is an agent skill for solo builders who implement libraries and packages from RFCs or formal specs and need more than vibes-based review. It bridges indexion’s quantitative alignment tooling with spec-driven development agent flows (cc-sdd, Codex, Claude) so you can draft SDD requirements from upstream documents, align them with what was actually built, and re-run verification until drift is acceptable. The skill documents project bootstrap with cc-sdd in skills mode, language alignment pitfalls, and when to invoke the loop—especially after codex spec-impl and before merging. It fits indie teams treating specs as contracts: you get measurable conformance signals plus a clear handoff for agents to fix gaps, and you can wire the same checks into CI so production-bound branches do not silently diverge from the source of truth.509installs7Indexion Wikiindexion-wiki is an agent-oriented runbook for the indexion project wiki lifecycle at .indexion/wiki/. Solo and indie builders using Claude Code, Cursor, or similar agents install it when they adopt indexion’s manifest-driven wiki instead of loose Markdown notes. The skill explains that a page is four synchronized artifacts: the Markdown file, wiki.json entry, search indexes (vectors.db, search-sections.json, tfidf-vocabulary.json), and log.json audit trail—only indexion wiki pages add and wiki pages update keep all four aligned. It covers creating pages, ingesting upstream source changes, updating stale content, linting structure, reconciling drift against code, and rebuilding indexes. Use it whenever an agent navigates or maintains the wiki so search, provenance, and audit history stay trustworthy. Skip it if the project does not use indexion; for those repos, ordinary docs skills suffice.508installs8Indexion Agent Orientindexion-agent-orient teaches a repeatable workflow for coding agents that tend to edit the first matching path without understanding ownership. Before implementation in an ambiguous area, you verify a current indexion binary, run agent orient to materialize a structure brief from the full orientation map, and inject that brief into agent instructions. The skill stresses provenance: owner profiles lean on package docs and owner-specific wiki pages, while wide wiki pages may inform context but must not over-claim ownership. Display limits on --limit do not shrink the evidence base used for inference. Solo builders maintaining large monorepos or unfamiliar forks benefit most when every agent session would otherwise rediscover the same landmines. It complements indexion’s wiki and reconcile commands rather than replacing reading the code.505installs9Indexion IdentityIndexion Identity is an agent skill that walks solo builders through detecting name and content drift in an existing codebase using the Indexion CLI. You start with `indexion identity audit`, optionally export a machine-readable queue, and interpret each row as evidence to verify—not automatic proof of a bad name. The skill explains how expected summaries from paths and scopes compare to actual summaries from declarations, docs, and module notes, and when files should stay evidence-thin until inspection. Before editing, you confirm symbols with `indexion doc graph`, semantic name grep, and ripgrep so renames, moves, folder changes, and splits stay safe. It fits indie teams who rename often and want agent-guided audits instead of ad-hoc string replaces. Install it when names have outlived refactors or when planning structural cleanup before release.505installs