ksh999000/Claude-Code-Game-Studios
1
0
Fork 0
mirror of https://github.com/Donchitos/Claude-Code-Game-Studios.git synced 2026-10-16 01:37:05 +00:00
Code Issues Packages Projects Releases Wiki Activity

Release 1.1.2 — the default path works end to end

Browse Source 
Fix release for 1.1.1: the minimal path end to end, engine commands that
run on Godot, Unity and Unreal, hook warnings you can see, and verdicts
that say when they could not check. Details in CHANGELOG.md; upgrade
steps in UPGRADING.md (v1.1.1 → v1.1.2).

Fixes #67
Fixes #131

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01WuJn2FdynEybfL5KC697nJ
This commit is contained in:
Donchitos
2026-09-29 21:32:19 +10:00
co-authored by Claude Opus 5.5
parent 7ed2c3e9c4
commit b21fa0f7f2
383 changed files with 22888 additions and 10618 deletions
Download Patch File Download Diff File Expand all files Collapse all files
CCGS Skill Testing Framework/templates/agent-test-spec.md
+13 -16
View File
@@ -1,8 +1,4 @@
# Agent Spec: [agent-name]
> **Tier**: [directors | leads | specialists | godot | unity | unreal | operations | creative]
> **Category**: [director | lead | specialist | engine | operations | creative]
> **Spec written**: [YYYY-MM-DD]
# Agent Test Spec: [agent-name]
## Agent Summary
@@ -15,10 +11,11 @@ delegates vs. handles directly. Include which gates it triggers (if any).]
---
## Static Assertions
## Static Assertions (Structural)
- [ ] Agent file exists at `.claude/agents/[name].md`
- [ ] Frontmatter has `name`, `description`, `model`, `tools` fields
- [ ] Model tier is `[opus | sonnet | haiku | inherit]` — frontmatter `model:` reads exactly that alias (tiers: `.claude/docs/model-tiers.md`)
- [ ] Domain clearly stated
- [ ] Escalation path documented
- [ ] Does not make decisions outside its domain
@@ -45,7 +42,7 @@ delegates vs. handles directly. Include which gates it triggers (if any).]
- [ ] Output format matches expected structure
- [ ] Collaborative protocol followed (ask → draft → approve)
**Case Verdict**: PASS / FAIL / PARTIAL
**Case Verdict**: PASS / FAIL / PARTIAL / NOT ASSESSED
---
@@ -65,7 +62,7 @@ delegates vs. handles directly. Include which gates it triggers (if any).]
- [ ] Agent declines and redirects (does not silently handle cross-domain work)
- [ ] Correct agent named in redirect
**Case Verdict**: PASS / FAIL / PARTIAL
**Case Verdict**: PASS / FAIL / PARTIAL / NOT ASSESSED
---
@@ -78,16 +75,16 @@ delegates vs. handles directly. Include which gates it triggers (if any).]
- [gate ID: e.g., CD-PHASE-GATE]
**Expected behavior**:
1. Agent reads the relevant documents
2. Agent produces a PASS / CONCERNS / FAIL verdict
3. Agent does not auto-advance on CONCERNS or FAIL
1. Agent reads the gate's definition file, then the documents it names
2. Agent opens with `[GATE-ID]: [word]`, the word taken from that gate's **Verdicts** line
3. Agent returns the verdict to the calling skill and does not act on it itself
**Assertions**:
- [ ] Verdict keyword present in output (PASS, CONCERNS, FAIL)
- [ ] First line is `[GATE-ID]: [word]` with a word from the invoked gate's Verdicts line — or `NOT ASSESSED`, naming the input, when an input the gate names was not given
- [ ] Reasoning provided for verdict
- [ ] On CONCERNS/FAIL: work is blocked, not silently continued
- [ ] Agent does not carry on with the reviewed work past its own verdict — the calling skill, not the agent, blocks on the gate's failure word (CONCERNS is not blocking; the skill takes it to the user)
**Case Verdict**: PASS / FAIL / PARTIAL
**Case Verdict**: PASS / FAIL / PARTIAL / NOT ASSESSED
---
@@ -108,7 +105,7 @@ delegates vs. handles directly. Include which gates it triggers (if any).]
- [ ] Correct escalation path followed
- [ ] No unilateral cross-domain changes made
**Case Verdict**: PASS / FAIL / PARTIAL
**Case Verdict**: PASS / FAIL / PARTIAL / NOT ASSESSED
---
@@ -130,7 +127,7 @@ delegates vs. handles directly. Include which gates it triggers (if any).]
- [ ] Result is scoped to the sub-task, not expanded beyond it
- [ ] Output format suitable for parent agent consumption
**Case Verdict**: PASS / FAIL / PARTIAL
**Case Verdict**: PASS / FAIL / PARTIAL / NOT ASSESSED
---
CCGS Skill Testing Framework/templates/skill-test-spec.md
+34 -19
View File
@@ -1,8 +1,4 @@
# Skill Spec: /[skill-name]
> **Category**: [gate | review | authoring | readiness | pipeline | analysis | team | sprint | utility]
> **Priority**: [critical | high | medium | low]
> **Spec written**: [YYYY-MM-DD]
# Skill Test Spec: /[skill-name]
## Skill Summary
@@ -10,13 +6,13 @@
---
## Static Assertions
## Static Assertions (Structural)
These should pass before any behavioral testing:
Checked against the SKILL.md by `/skill-test spec` — no fixture needed.
- [ ] Frontmatter has all required fields (`name`, `description`, `argument-hint`, `user-invocable`, `allowed-tools`)
- [ ] 2+ phase headings found
- [ ] At least one verdict keyword present (`PASS`, `FAIL`, `CONCERNS`, `APPROVED`, `BLOCKED`, `COMPLETE`, `READY`)
- [ ] Contains the skill's verdict keywords — any that `/skill-test static` Check 3 accepts (gate or review verdicts, go/no-go verdicts, or a severity scale), plus `NOT ASSESSED` when anything depends on the verdict; none needed when the skill outputs an artifact rather than a judgement
- [ ] If `allowed-tools` includes Write/Edit: `"May I write"` language present
- [ ] Next-step handoff section present at end
@@ -27,8 +23,8 @@ These should pass before any behavioral testing:
[Describe which director gates this skill triggers (if any), and under what review mode conditions.]
- **Full mode**: [gates triggered — e.g., CD-PHASE-GATE, TD-PHASE-GATE, PR-PHASE-GATE, AD-PHASE-GATE]
- **Lean mode**: [phase gates only — e.g., CD-PHASE-GATE only, or none]
- **Solo mode**: [no gates — skill runs without director review]
- **Lean mode**: [phase gates only — for `/gate-check`, the panel `modes.workflow` sets (PR only at `minimal`, TD + PR at `standard`, all four at `full`); other skills skip their non-phase gates, each noted `[GATE-ID] skipped — Lean mode`]
- **Solo mode**: [no gates — skill runs without director review; each skip is noted, e.g. `[GATE-ID] skipped — Solo mode`]
- **N/A**: [if this skill never triggers gates, explain why]
---
@@ -51,7 +47,7 @@ These should pass before any behavioral testing:
- [ ] [Assertion 2]
- [ ] [Assertion 3]
**Case Verdict**: PASS / FAIL / PARTIAL
**Case Verdict**: PASS / FAIL / PARTIAL / NOT ASSESSED
---
@@ -66,11 +62,11 @@ These should pass before any behavioral testing:
3. [Skill does NOT proceed]
**Assertions**:
- [ ] Skill stops early and does not produce output
- [ ] Skill stops before producing the artifact and reports BLOCKED or NOT ASSESSED, naming the missing input
- [ ] Correct error/block message displayed
- [ ] No files written without user approval
**Case Verdict**: PASS / FAIL / PARTIAL
**Case Verdict**: PASS / FAIL / PARTIAL / NOT ASSESSED
---
@@ -87,7 +83,7 @@ These should pass before any behavioral testing:
- [ ] [Mode-specific assertion]
- [ ] [Output differs correctly from Case 1]
**Case Verdict**: PASS / FAIL / PARTIAL
**Case Verdict**: PASS / FAIL / PARTIAL / NOT ASSESSED
---
@@ -103,7 +99,7 @@ These should pass before any behavioral testing:
- [ ] [Edge case handled without crash or silent failure]
- [ ] [Correct output or message]
**Case Verdict**: PASS / FAIL / PARTIAL
**Case Verdict**: PASS / FAIL / PARTIAL / NOT ASSESSED
---
@@ -119,11 +115,30 @@ These should pass before any behavioral testing:
**Assertions**:
- [ ] In full mode: [specific gates spawn]
- [ ] In lean mode: [phase gates only, or skip]
- [ ] In solo mode: no director gates spawn
- [ ] Skill does not auto-advance past a CONCERNS or FAIL verdict
- [ ] In lean mode: [phase gates only, at the width `modes.workflow` sets — or each gate skipped with `[GATE-ID] skipped — Lean mode`]
- [ ] In solo mode: no director gates spawn, and the output names each skip
- [ ] Skill does not auto-advance past a CONCERNS, FAIL or NOT ASSESSED verdict
**Case Verdict**: PASS / FAIL / PARTIAL
**Case Verdict**: PASS / FAIL / PARTIAL / NOT ASSESSED
---
### Case 6: Could Not Assess — [brief name]
**Fixture**:
- [an input the skill needs is missing, unreadable, empty or placeholder-only — or a check it must run cannot run]
**Expected behavior**:
1. [Skill detects that part of its scope could not be evaluated]
2. [Skill reports NOT ASSESSED, naming what it could not assess and why]
3. [Skill does NOT resolve the unknown to a pass]
**Assertions**:
- [ ] Verdict is NOT ASSESSED — never the pass value — and names the missing or unreadable input
- [ ] A failure found alongside it still takes the failure verdict (NOT ASSESSED ranks above the pass value, below the failure values)
- [ ] The check that could not run is named in the output, not silently omitted
**Case Verdict**: PASS / FAIL / PARTIAL / NOT ASSESSED
---