docs: reconcile skill count to 10 after research-grounded

.claude/skill-context.md

@@ -9,12 +9,12 @@ audits the skill collection itself.
 ## repo
 
 - name: techne
-- package_root: `plugins/techne/skills/` (9 skills as `SKILL.md` + supporting markdown), `plugins/techne/_shared/` (canonical glossaries shared across skills), `superpowers/` (plans + specs), `scripts/` (validation helpers)
+- package_root: `plugins/techne/skills/` (10 skills as `SKILL.md` + supporting markdown), `plugins/techne/_shared/` (canonical glossaries shared across skills), `superpowers/` (plans + specs), `scripts/` (validation helpers)
 - language: Markdown (skill bodies) + Python (validation snippets in workflows) + Bash (structural-check scripts)
 - cli_entrypoint: none — skills are invoked from the consumer's Claude Code via `/plugin install techne@techne` then `/techne:<skill>`. The repo itself is `package = false` in `pyproject.toml`.
 - runner_module: no Python runner; CI gates run inline in `.github/workflows/validate.yml`.
 - default_branch: `main`
-- has: 9 skills (`audit`, `auto-commit`, `ci-audit`, `deslop`, `docs-site`, `docsync`, `reslop`, `sisters`, `theoros`), plugin manifest at `plugins/techne/.claude-plugin/plugin.json`, marketplace manifest at `.claude-plugin/marketplace.json`, Zensical-powered docs site, no docker, no frontend
+- has: 10 skills (`audit`, `auto-commit`, `ci-audit`, `deslop`, `docs-site`, `docsync`, `research-grounded`, `reslop`, `sisters`, `theoros`), plugin manifest at `plugins/techne/.claude-plugin/plugin.json`, marketplace manifest at `.claude-plugin/marketplace.json`, Zensical-powered docs site, no docker, no frontend
 
 ## audit
 
@@ -86,7 +86,7 @@ Source of truth for skill-level claims:
 
 - **Skill descriptions:** `plugins/techne/skills/<name>/SKILL.md` frontmatter `description:` field is the canonical one-line summary surfaced in the plugin registry; README and `docs/skills/*.md` cross-references must match.
 - **Marketplace metadata:** `.claude-plugin/marketplace.json` lists each skill; the descriptions there must align with the SKILL.md frontmatter.
-- **Skill count:** `find plugins/techne/skills -mindepth 1 -maxdepth 1 -type d | wc -l` is the canonical count. README + docs claims about "9 skills" or similar must trace here.
+- **Skill count:** `find plugins/techne/skills -mindepth 1 -maxdepth 1 -type d | wc -l` is the canonical count. Count + skill-list claims in README, ROADMAP, IMPL, **and this file** (e.g. "10 skills") must all trace here — adding a skill touches every one.
 
 Any quantitative or list-shape claim not traceable to one of those is slop.
 

IMPL.md

@@ -24,12 +24,12 @@ Next natural pickup: any of the queued skills from ROADMAP
 
 ## Skill collection state
 
-Nine skills shipped as of 2026-05-23, four catalog dimensions:
+Ten skills shipped as of 2026-05-29, four catalog dimensions:
 
 | Dimension | Skills |
 | --- | --- |
 | **Audit** | `audit`, `ci-audit` |
-| **Drift** | `docsync`, `docs-site`, `sisters` |
+| **Drift** | `docsync`, `docs-site`, `research-grounded`, `sisters` |
 | **Hygiene** | `auto-commit`, `deslop`, `reslop` |
 | **Observation** | `theoros` |
 

ROADMAP.md

@@ -55,11 +55,11 @@ gating. Open coverage gaps:
 
 Skills are added when a pattern proves itself across multiple sisters
 (n≥2). Skills are deleted or merged when their domain collapses into a
-larger sibling. The current set (9 skills, 4 catalog dimensions —
-audit, drift, hygiene, observation) is stable; the next addition will
-likely be `narrative-coherence` or `positioning` once the
-2026-05-21 audit-of-audit drift class shows up a second time in the
-wild.
+larger sibling. The current set (10 skills, 4 catalog dimensions —
+audit, drift, hygiene, observation) is stable. `research-grounded`
+(the 2026-05-21 audit-of-audit item #3) shipped 2026-05-29 on direct
+request; `narrative-coherence` / `positioning` (items #2 / #1) remain
+queued for when that drift class recurs.
 
 ---