docs: align every roster introduction on the eight Participant CLIs

[dichovsky]

Closes #26

What changed and why

src/participants.ts:7-16 and src/platforms/registry.ts declare eight Participant CLIs, but every top-level introduction understated the roster — and each understated it differently. Per the authority order in docs/README.md, the registry is authoritative and the restatements are the bug, so this is a documentation-and-metadata correction with no behavior change.

README.md — three sites, all at five:

• Intro (README.md:7-12): "Tools such as Claude Code, Codex CLI, Gemini CLI, GitHub Copilot CLI, and Little Coder each run as..." → "The eight supported tools — Claude Code, Codex CLI, Gemini CLI, GitHub Copilot CLI, Antigravity CLI, Pi CLI, Little Coder, and opencode CLI — each run as...".
• "Not locked to one vendor" bullet: gains Antigravity CLI, Pi CLI, and opencode CLI.
• Architecture diagram: the subgraph declared A1–A5, but the edge into BIN wired only A1 & A2 & A3 & A4 — so even the already-declared Little Coder node was orphaned. Now A1–A8 in registry order, with the edge covering all eight.

CLAUDE.md / AGENTS.md — both listed seven, omitting Little Coder. Both now read "the eight Participant CLIs listed in src/participants.ts (…)", adding an authoritative pointer alongside the inline names. The two files remain byte-identical in the changed span.

package.json — description now names all eight; keywords gain antigravity, pi, and opencode, which were search-invisible on npm despite being shipped. No version bump, no other field touched. docs/ is not in files, and this metadata reaches the registry only at the next publish.

docs/design/srs.md — §1.1 and §1.3.1 listed four while Appendix A (srs.md:1172) already said "The eight Participant CLIs" and the Dependencies list (srs.md:1176-1178) already named all eight. Both now reference Appendix A rather than restating, per the issue's stated preference. FR-G07 (srs.md:511-514) gains Antigravity in the /crew group — src/platforms/antigravity.ts:28-30 returns `/crew ${role} ${id}`, and Appendix E (srs.md:1387) already grades FR-G07 as "Lists eight per-platform invocations". FR-G07 keeps its explicit enumeration because it is normative and Appendix E grades it specifically on listing eight. No FR-*/NFR-* id was added or renumbered.

docs/design/product-spec.md — line 70 promised setup generators for four, against its own gate table (line 126) saying "each of the eight Participant CLIs". Now aligned.

Already-correct sites were deliberately left alone: srs.md:1172, srs.md:1176-1178, srs.md:1387, and product-spec.md:126.

No src/, tests/, bin/, web/, or docs-site/ file is touched. tests/unit/docs-facts.test.ts extracts its roster from PARTICIPANT_TARGETS into docs-site/generated/facts.json and never reads this prose, so no fact regeneration is required.

Verification

Docs-and-metadata-only change: no tests added, claiming the docs-only exemption in .github/ISSUE_TRIAGE.md. The full gate was still run, under Node 24.18.0 (the CI version).

Command	Result
npm run typecheck	pass — root + web/ + docs-site/ projects
npm run lint	pass — no output
npm run format:check	pass — "All matched files use Prettier code style"
npm run build	pass
npm run test:coverage	fail (exit 1) — 1395 passed, 3 skipped, 2 pre-existing environment failures

Both failures are in tests/integration/commands/doctor.test.ts ("degrades a raw filesystem read failure in project roles instead of aborting doctor" and "degrades unreadable roles and teams directories to whole-listing warnings"). They reproduce identically on an unmodified tree at d311767 — verified by stashing the change and re-running the file. Root cause is environmental: the sandbox runs as uid 0, so the tests' chmod 000 does not actually make the directories unreadable and the expected INVALID_CONFIG warnings never fire.

Because vitest suppresses the coverage table on a failing run, the 95% thresholds were not evaluated locally — not on this branch and not on the clean tree. No src/** or bin/** file changed, so coverage is unchanged from main. This PR's CI run on GitHub-hosted runners is the authority for the gate.

Rebased onto origin/main @ d311767 immediately before pushing (a no-op — the base had not moved) and the gate was re-run afterward.

Related open PRs

• docs: trace crew team resume with SRS requirements #49 — docs: trace crew team resume with SRS requirements. Touches docs/design/srs.md, the same file this PR edits, but in disjoint regions: docs: trace crew team resume with SRS requirements #49 adds FR-U39–FR-U42 in group U and edits §1.3.2, the v1-recount line, and the Appendix E section-default row; this PR edits §1.1, §1.3.1, and FR-G07. Whichever merges second may need a trivial rebase, but the hunks do not overlap.

Out of scope

Deliberately untouched, each tracked separately:

• docs-site prose roster drift — docs-site prose drift: lifecycle omits abandoned, deferred-seam roster, glossary count #42.
• The wider CLAUDE.md/AGENTS.md command-annotation drift — CLAUDE.md/AGENTS.md command annotations have drifted from the actual scripts and gates #41.
• README.md:8 says "GitHub Copilot CLI" while every other surface uses the bare "Copilot CLI". That is naming-consistency drift, not roster drift; the existing wording was preserved.

---

Generated by Claude Code

[Copilot]

Pull request overview

Aligns top-level documentation and npm package metadata with the authoritative eight-Participant-CLI roster defined in src/participants.ts / the platforms registry, resolving documented roster drift without changing runtime behavior.

Changes:

• Update README.md introductions and the Mermaid architecture diagram to reflect all eight Participant CLIs.
• Update onboarding docs (CLAUDE.md, AGENTS.md) and specs (docs/design/srs.md, docs/design/product-spec.md) to reference/reflect the eight-CLI roster, including adding Antigravity to FR-G07’s normative invocation list.
• Refresh package.json description and keywords so npm metadata matches the shipped roster.

Reviewed changes

Copilot reviewed 6 out of 6 changed files in this pull request and generated no comments.

Show a summary per file

File	Description
README.md	Updates intro/bullets and architecture diagram to list all eight Participant CLIs.
package.json	Expands description + keywords to include Antigravity, Pi, and opencode for npm discoverability.
docs/design/srs.md	Removes undercounted restatements by pointing to Appendix A; updates FR-G07 to include Antigravity.
docs/design/product-spec.md	Updates v1 capability scope bullet to include all eight Participant CLIs.
CLAUDE.md	Corrects the “What this is” roster to eight and points to src/participants.ts.
AGENTS.md	Mirrors CLAUDE.md roster correction and src/participants.ts pointer.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[dichovsky]

Routine review. Roster per .github/ISSUE_TRIAGE.md: general code reviewer (always) plus the docs-only prose-vs-code verification in place of the TypeScript reviewer. No security reviewer — the diff touches none of src/store/, src/process.ts, src/which.ts, src/fs-safe.ts, src/setup/, or the Launcher.

CRITICAL: none. HIGH: none. No fix pass triggered.

Verified

• Roster and order. All six enumerations (README prose ×2, README mermaid A1–A8, CLAUDE.md/AGENTS.md, package.json, product-spec.md:70-71) are exactly the eight ids of src/participants.ts:7-16 in the canonical order of src/platforms/registry.ts:20-29. Display names match docs/design/setup-integration.md §4.1–§4.8, including the deliberately lowercase "opencode CLI" and bare "Little Coder".
• FR-G07 groupings are exactly right, checked against every invocation(): /crew — claude.ts:34, gemini.ts:35, antigravity.ts:29, pi.ts:51, little-coder.ts:40, opencode.ts:33; $crew — codex.ts:30; /agent (select crew), then: … — copilot.ts:34. The cited verification test really does assert all eight (tests/unit/platforms.test.ts:114-124). This reconciles FR-G07 with Appendix E's existing grade at srs.md:1390.
• "Setup generators for each of the eight" does not over-promise. All eight modules implement render(), src/setup/index.ts:174 writes it, and cli-contract.md:37 accepts all eight ids.
• Mermaid parses. A5–A8 declared and consumed only by the single &-chained edge; no dangling or duplicate ids. The edge previously wired only A1–A4, orphaning the already-declared A5 — both halves are now consistent.
• Authority direction is correct. Appendix A (srs.md:1175, :1179-1181) and Appendix E (srs.md:1390) already said eight; §1.1, §1.3.1 and FR-G07 said four/seven. The lower/older prose was fixed to match the registry, adding no obligation the code does not already meet. No FR-*/NFR-* id added or renumbered. No accepted ADR states a roster this contradicts (ADR-0016's five-CLI list is a scoped mid-2026 survey, and its own Update 2026-07-14 already adds pi-cli/opencode-cli).
• Scope. git diff --check clean; rewrapping is confined to paragraphs that had to change. CLAUDE.md and AGENTS.md remain byte-identical in the changed span. package.json version untouched; docs-facts.test.ts reads only name/version/engines/deps from it, so no fact regeneration is needed.
• CI is green on GitHub-hosted runners (ubuntu / node 24.18.0, plus publish dry-run — which matters here because package.json changed). That supersedes the two local doctor failures reported in the PR body, which were a uid-0 sandbox artifact.

MEDIUM

1. docs/design/architecture.md:458-459 — the same understatement FR-G07 just fixed, still present one authority level down: "Claude and Gemini use /crew, Codex uses $crew, and in Copilot you run /agent…" omits Antigravity, Pi, Little Coder, and opencode, all of which return /crew. This disagreement is pre-existing — architecture.md already omitted Pi, Little Coder, and opencode before this PR — so it is not something the change introduced, and .github/ISSUE_TRIAGE.md's scope rule directs a distinct problem in an unnamed file to a new issue rather than into this diff. Filed as a follow-up rather than folded in.

2. web/view-model.ts:141-161 — ENGINE_META enumerates seven of eight; little-coder is absent, so engineMeta() (line 170) falls through to the unknown-engine badge and the Console renders the raw id. A genuine code-side roster gap, correctly outside a docs-only PR. Filed separately.

3. tests/tools/release-smoke.test.ts:15 — maintainer instruction still says "the five Participant CLIs / backends installed" while the recorder iterates PARTICIPANT_TARGETS + BACKEND_TARGETS (eight plus two). Same class of drift, unnamed file; filed with (1).

LOW

4. README.md:26-28 — the rewritten bullet keeps the inherited phrase "…and locally hosted models can all join the same Crew". CONTEXT.md:181 resolves exactly this ambiguity: Ollama and LM Studio are Setup Targets and Model Backends, never Participants, and a Crew contains Agents (CONTEXT.md:155). A locally hosted model does not join a Crew; a Participant CLI pointed at one does. Wording is pre-existing but sits on a line this PR rewrote.

5. CLAUDE.md:7-9 / AGENTS.md:7-9 — the new sentence points at src/participants.ts for names, but that module is "intentionally only an id vocabulary" (src/participants.ts:1-6); display names and canonical order live in PARTICIPANT_TARGETS (src/platforms/registry.ts:20-29). Count is right; the pointer is slightly off-target.

6. docs/design/srs.md:53, :88 — the two new #appendix-a--assumptions-dependencies-and-tbd-items links do not resolve, because the heading's trailing (§9.5.7) contributes -957 to the GitHub slug. Pre-existing convention, not introduced here: the TOC at srs.md:30 and the link at srs.md:1034 have the identical defect, and no link checker runs in CI. The PR is internally consistent; flagging only so it is not misattributed.

7. Durability — tests/unit/docs-facts.test.ts guards only docs-site/generated/facts.json; its own docstring records that prose roster drift "has already happened once" (commit 8820be7). This PR is the second manual reconciliation and adds more hand-maintained restatements. Extending the fact-extraction guard to the root docs would prevent a third recurrence — worth a follow-up, out of scope here.

Nothing at MEDIUM or LOW blocks this change. Findings 1–3 are filed as separate issues per the scope rule; 4–7 are recorded for a maintainer's judgment.

---

Generated by Claude Code