boost-skill-cli 1.0.177__tar.gz → 1.0.178__tar.gz

boost_skill_cli-1.0.178/.markdownlint-cli2.jsonc

@@ -0,0 +1,32 @@
+// markdownlint-cli2 — consistency for every contributor-facing Markdown
+// surface (README, CLAUDE.md, CONTRIBUTING, SECURITY, docs/*.md, examples,
+// style). Run locally with:
+//   npx --yes markdownlint-cli2@0.18.1
+// CI: .github/workflows/markdownlint.yml (path-filtered).
+{
+  "globs": ["**/*.md"],
+  "ignores": [
+    ".venv/**",
+    "node_modules/**",
+    "mutants/**",
+    ".eval-home/**",
+    // Agent-harness config (skills/commands), not contributor documentation.
+    ".claude/**",
+    // Roadmap item bodies are HTML-in-Markdown card copy rendered by
+    // build_roadmap.py, not standalone documents — heading/inline-HTML
+    // document rules don't apply to them.
+    "docs/roadmap/items/**"
+  ],
+  "config": {
+    // Long lines are fine: prose wraps in the viewer, and tables/links often
+    // can't be split without hurting readability.
+    "MD013": false,
+    // Inline HTML is deliberate (badges, <details>, <img> sizing in README).
+    "MD033": false,
+    // The README opens with a centered logo block, not an H1.
+    "MD041": false,
+    // "boost — Homebrew for AI coding skills." style headings end with
+    // punctuation on purpose; it's the product tagline idiom.
+    "MD026": false
+  }
+}

{boost_skill_cli-1.0.177 → boost_skill_cli-1.0.178}/CLAUDE.md

@@ -33,6 +33,7 @@ is relaxed in the gate (`--regression-eps 1`) so upstream repo drift can't flake
 it. Edit `taps.txt` → regenerate `tests/eval/baseline.json`. It runs in CI's
 `lint` job (pure-stdlib BM25, no `ANTHROPIC_API_KEY`; needs network to tap). The
 opt-in evals stay out of `check` and all degrade cleanly:
+
 - `make eval-ai` / `eval-rec` — key-gated LLM evals (Tier 2a rerank / 2b recommend).
 - `make eval-stats` — Tier 1b `ranx` paired-t-test between engines (`--stats`).
 - `make eval-explain` — Tier 2c `ragas` faithfulness for `boost explain`; needs

{boost_skill_cli-1.0.177 → boost_skill_cli-1.0.178}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: boost-skill-cli
-Version: 1.0.177
+Version: 1.0.178
 Summary: boost — Homebrew for AI coding skills
 Author: Jonathan Reyes
 License: GNU GENERAL PUBLIC LICENSE
@@ -818,7 +818,7 @@ installed skills into one canonical store, and then symlinks them out to
 whichever agents you've got — all of it tracked in a lock file so state
 stays reproducible.
 
-```
+```text
 GitHub registries  ──boost update──▶  ~/.boost/repos/    (shallow clones)
                                      ~/.boost/cache/    (JSON catalogs)
                    ──boost install─▶  ~/.agents/skills/ (canonical store)

{boost_skill_cli-1.0.177 → boost_skill_cli-1.0.178}/README.md

@@ -102,7 +102,7 @@ installed skills into one canonical store, and then symlinks them out to
 whichever agents you've got — all of it tracked in a lock file so state
 stays reproducible.
 
-```
+```text
 GitHub registries  ──boost update──▶  ~/.boost/repos/    (shallow clones)
                                      ~/.boost/cache/    (JSON catalogs)
                    ──boost install─▶  ~/.agents/skills/ (canonical store)

{boost_skill_cli-1.0.177 → boost_skill_cli-1.0.178}/boost_cli/_version.py

@@ -18,7 +18,7 @@ version_tuple: tuple[int | str, ...]
 commit_id: str | None
 __commit_id__: str | None
 
-__version__ = version = '1.0.177'
-__version_tuple__ = version_tuple = (1, 0, 177)
+__version__ = version = '1.0.178'
+__version_tuple__ = version_tuple = (1, 0, 178)
 
 __commit_id__ = commit_id = None

{boost_skill_cli-1.0.177 → boost_skill_cli-1.0.178}/boost_skill_cli.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: boost-skill-cli
-Version: 1.0.177
+Version: 1.0.178
 Summary: boost — Homebrew for AI coding skills
 Author: Jonathan Reyes
 License: GNU GENERAL PUBLIC LICENSE
@@ -818,7 +818,7 @@ installed skills into one canonical store, and then symlinks them out to
 whichever agents you've got — all of it tracked in a lock file so state
 stays reproducible.
 
-```
+```text
 GitHub registries  ──boost update──▶  ~/.boost/repos/    (shallow clones)
                                      ~/.boost/cache/    (JSON catalogs)
                    ──boost install─▶  ~/.agents/skills/ (canonical store)

{boost_skill_cli-1.0.177 → boost_skill_cli-1.0.178}/boost_skill_cli.egg-info/SOURCES.txt

@@ -2,6 +2,7 @@
 .gitignore
 .gitleaks.toml
 .htmlvalidate.json
+.markdownlint-cli2.jsonc
 .pre-commit-config.yaml
 CLAUDE.md
 CONTRIBUTING.md

{boost_skill_cli-1.0.177 → boost_skill_cli-1.0.178}/docs/DEBUGGING.md

@@ -25,7 +25,7 @@ a run that looked completely normal.
 
 Every invocation appends to a rotating log file:
 
-```
+```text
 ~/.boost/logs/boost.log        # current file (rotates at ~1 MB)
 ~/.boost/logs/boost.log.1      # … up to 3 older files (~4 MB ceiling total)
 ```
@@ -47,7 +47,7 @@ Every invocation is bookended by two lines — an `invoke:` at the start and a
 the trail doubles as a lightweight timing record. A failing run logs its `done:`
 line at `WARNING`, so it stands out when you scan the log:
 
-```
+```text
 … INFO    boost: invoke: boost install brainstorming
 … INFO    boost: done: boost install brainstorming -> rc=0 in 214ms
 … WARNING boost: done: boost install missing -> rc=1 in 48ms
@@ -93,7 +93,7 @@ print a clean one-line message and hint), boost:
 
 A crash report bundles everything needed to reproduce or file a bug:
 
-```
+```text
 boost crash report
 ==================
 time:     2026-07-18T05:23:14Z

{boost_skill_cli-1.0.177 → boost_skill_cli-1.0.178}/docs/rag-architecture.md

@@ -8,7 +8,8 @@ through the MCP server — without breaking boost's zero-dependency posture.
 
 ## 1. Goals & non-goals
 
-**Goals**
+### Goals
+
 1. **Natural-language retrieval.** An agent asks "best skill for testing a React app" and boost returns a
    semantically-ranked shortlist with reasons — not just keyword hits.
 2. **Full-content search.** Rank over the *whole* skill/rule/workflow body, not only the frontmatter
@@ -18,7 +19,8 @@ through the MCP server — without breaking boost's zero-dependency posture.
 4. **MCP-hub extensibility.** Position the MCP server to later call out to GitHub, a database, and other
    dependencies as drop-in tools.
 
-**Non-goals**
+### Non-goals
+
 - Forcing FastAPI/django (the MCP transport is stdio JSON-RPC, not HTTP — a web framework has nothing to
   serve here).
 - Making a vector database or `langchain` a *required* runtime dependency.
@@ -124,6 +126,7 @@ and `commits`.
 **No third-party imports. Stays green through `make check`.**
 
 ### 6a. Engine — `boost_cli/core/rag.py`
+
 Pure-Python **BM25** over the chunk corpus (BM25 subsumes TF-IDF and handles document-length
 normalization, which matters when bodies vary from 3 lines to 300). Sparse term vectors are a legitimate
 "vector" retrieval — we get semantic *recall breadth* from full-text coverage, and the LLM rerank
@@ -133,6 +136,7 @@ Scoring is standard BM25 (`k1≈1.2`, `b≈0.75`). At ~9k items / tens of thousa
 dict-of-postings scan — comfortably sub-second per query in Python.
 
 ### 6b. Two-stage pipeline — upgrade `boost_search` / `catalog.search` **in place**
+
 No parallel MCP tool (per decision). The upgraded flow:
 
 1. **Retrieve** top-N (~60) candidate entries via `Retriever.query(task, k=60, kind=?)`.
@@ -148,12 +152,14 @@ No parallel MCP tool (per decision). The upgraded flow:
 `cmd_search` (`_ai_rank`, `discovery.py:150`) is folded into / superseded by this path.
 
 ### 6c. New command — `boost reindex`
+
 Register `("reindex", "find", "discovery", "Build/refresh the full-content search index")` in the
 `COMMANDS` table (`boost_cli/cli.py:48`) and add `cmd_reindex(argv)` in `discovery.py`. It calls
 `rag.build(catalog.all_entries())`, honoring commit-keyed incremental refresh, and prints doc/entry
 counts. The search path also lazily triggers a refresh when it detects a tap commit newer than the index.
 
 ### 6d. Tests & gates (Phase 1)
+
 - `tests/unit/test_rag.py` — chunking, BM25 scoring, incremental refresh, degradation paths. Because
   `rag.py` lives in `core/`, it is **mutation-gated at 80%** (`scripts/mutation_gate.py`) — write tests
   that pin the scoring math and the fallback branches.
@@ -179,6 +185,7 @@ rag = [ "..." ]   # dense-embedding + vector-store stack
 ```
 
 A second `Retriever` implementation (`engine="dense"`) that:
+
 - **Embeds** each chunk into a dense vector — via a local model or an embeddings API over `urllib`
   (note: Anthropic exposes **no** embeddings endpoint; the API options are Voyage AI, which Anthropic
   recommends, or OpenAI). API keys follow the `core/ai.py` env-var convention.

{boost_skill_cli-1.0.177 → boost_skill_cli-1.0.178}/docs/roadmap/items/markdown-consistency-markdownlint-cli2.md

@@ -2,15 +2,15 @@
 id: markdown-consistency-markdownlint-cli2
 board: code
 section: docsite
-status: planned
+status: shipped
 category: Quality · Docs
 complexity: S
 impact: Low
 wow: 2
 note: every .md
 order: 7
-owner:
-pr:
+owner: loop/mdlint
+pr: 201
 title: Markdown consistency — <code>markdownlint-cli2</code>
 ---
 Lint every Markdown surface — README, <code>CLAUDE.md</code>,

{boost_skill_cli-1.0.177 → boost_skill_cli-1.0.178}/docs/roadmap.html

@@ -1605,7 +1605,7 @@ and the work stays visible.</p>
       </article>
 
       <article class="cap rcard">
-        <div class="head"><span class="pill planned">Planned</span><span class="cat">Quality · Docs</span></div>
+        <div class="head"><span class="pill shipped">Shipped</span><span class="cat">Quality · Docs</span></div>
         <h3>Markdown consistency — <code>markdownlint-cli2</code></h3>
         <p>Lint every Markdown surface — README, <code>CLAUDE.md</code>,
            <code>DEBUGGING.md</code>, <code>rag-architecture.md</code> — for

{boost_skill_cli-1.0.177 → boost_skill_cli-1.0.178}/style/README.md

@@ -11,7 +11,7 @@ gated behind `prefers-reduced-motion`.
 
 Open [`demo.html`](demo.html) to see the whole system on one page.
 
-```
+```text
 style/
 ├── boost.css     the design system — tokens, atmosphere, primitives, a11y
 ├── boost.js      two progressive enhancements: cursor glow + reveal-on-scroll