create-agentic-pdlc 2.2.1 → 2.3.0
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/.agentic-pdlc/SETUP_PROMPT.md +3 -4
- package/.agentic-pdlc/metrics/raw/2026-W18.jsonl +2 -0
- package/.agentic-pdlc/metrics/raw/2026-W21.jsonl +68 -0
- package/.agentic-pdlc/templates/.github/workflows/agent-trigger.yml +2 -2
- package/.agentic-pdlc/templates/.github/workflows/agentic-metrics.yml +16 -9
- package/.agentic-pdlc/templates/.github/workflows/ci.yml +14 -0
- package/.agentic-pdlc/templates/.github/workflows/pdlc-health-check.yml +0 -2
- package/.agentic-pdlc/templates/.github/workflows/project-automation.yml +1 -5
- package/.agentic-pdlc/templates/.github/workflows/qa-agent.yml +4 -4
- package/.agentic-pdlc/templates/AGENTS.md +24 -1
- package/.agentic-pdlc/templates/docs/pdlc.md +21 -13
- package/.agentic-setup-prompt.md +3 -4
- package/.agentic-setup.md +3 -4
- package/.github/workflows/agentic-metrics.yml +21 -11
- package/.github/workflows/pdlc-health-check.yml +0 -2
- package/.github/workflows/project-automation.yml +31 -7
- package/.github/workflows/qa-agent.yml +4 -4
- package/AGENTS.md +1 -1
- package/CLAUDE.md +51 -3
- package/SETUP.md +2 -0
- package/adapters/claude-code/skill.md +13 -10
- package/adapters/hooks/pdlc-stage-gate.sh +3 -3
- package/bin/cli.js +18 -7
- package/docs/flow.md +8 -21
- package/docs/pdlc.md +21 -13
- package/package.json +1 -1
- package/templates/.github/workflows/agent-trigger.yml +2 -2
- package/templates/.github/workflows/ci.yml +14 -0
- package/templates/.github/workflows/pdlc-health-check.yml +0 -2
- package/templates/.github/workflows/project-automation.yml +24 -6
- package/templates/.github/workflows/qa-agent.yml +6 -4
- package/templates/AGENTS.md +25 -2
- package/templates/docs/pdlc.md +21 -13
package/CLAUDE.md
CHANGED
|
@@ -1,9 +1,57 @@
|
|
|
1
|
+
# agentic-pdlc
|
|
2
|
+
|
|
3
|
+
**What it is:** npm CLI (`npx create-agentic-pdlc`) that installs a PDLC workflow for AI agent projects. Stack: Markdown + GitHub Actions YAML only — never complex Node/Python/Bash scripts.
|
|
4
|
+
|
|
5
|
+
**Workflow:** `stage:brainstorming` → `stage:detailing` → `spec:approved` → `stage:development` → PR → merge. Labels move the board automatically.
|
|
6
|
+
|
|
7
|
+
## Session Startup
|
|
8
|
+
|
|
9
|
+
Read: `AGENTS.md` (mandatory contract), `docs/pdlc.md` (board + labels — only when operating on the board).
|
|
10
|
+
Run: `gh issue list --state open --label "stage:development" --json number,title --jq '.[] | "#\(.number) \(.title)"'`
|
|
11
|
+
|
|
1
12
|
# PDLC Stage Gate
|
|
2
13
|
|
|
3
14
|
NEVER run `gh pr create` unless one of these is true:
|
|
4
|
-
- The linked issue has label `
|
|
15
|
+
- The linked issue has label `spec:approved`
|
|
5
16
|
- The branch name starts with `hotfix/`
|
|
6
17
|
|
|
7
|
-
|
|
18
|
+
NEVER edit files, create branches, or commit unless the linked issue has label `spec:approved` (set by human PM only) or the branch name starts with `hotfix/`.
|
|
19
|
+
|
|
20
|
+
Advance stages first: `brainstorming` → `detailing` → `approval` → (human adds `spec:approved`) → `development`
|
|
21
|
+
|
|
22
|
+
The PreToolUse hook will block `gh pr create` automatically if this rule is violated.
|
|
23
|
+
|
|
24
|
+
## Human-in-the-Loop
|
|
25
|
+
|
|
26
|
+
| Transition | Gate |
|
|
27
|
+
|---|---|
|
|
28
|
+
| → `stage:brainstorming` | Autonomous — apply immediately |
|
|
29
|
+
| → `stage:detailing` | Ask user — present problem summary + options, wait for choice |
|
|
30
|
+
| → `stage:approval` | **Autonomous** — agent completes spec end-to-end, advances without asking |
|
|
31
|
+
| `spec:approved` | **Human PM only** — agent waits; never adds this label |
|
|
32
|
+
| → `stage:development` | Human applies `spec:approved` label — that IS the gate |
|
|
33
|
+
|
|
34
|
+
**Detailing is fully autonomous.** Write the complete spec, add it to the issue, advance to `stage:approval` — no confirmation needed. Then **stop and wait** for human to add `spec:approved` before any implementation.
|
|
35
|
+
|
|
36
|
+
## Stage Transition Rules (non-negotiable)
|
|
37
|
+
|
|
38
|
+
MUST apply `stage:brainstorming` label immediately on starting work — before
|
|
39
|
+
reading any code, before invoking any skill. Then read context and present
|
|
40
|
+
problem summary + 2–3 solution options in a single message.
|
|
41
|
+
|
|
42
|
+
MUST NOT add `stage:detailing` label until the user has explicitly selected
|
|
43
|
+
an approach in the current conversation turn. Work done in a prior
|
|
44
|
+
planning session does NOT count as confirmation.
|
|
45
|
+
|
|
46
|
+
MUST NOT add `spec:approved` or any approval-trigger label under any
|
|
47
|
+
circumstances — these are set by the PM (human) only, after reviewing the spec
|
|
48
|
+
in the issue body. Adding them triggers irreversible automation (Jules dispatch,
|
|
49
|
+
board move).
|
|
50
|
+
|
|
51
|
+
MUST NOT add or remove `stage:development`, `spec:approved`, or `qa:*` labels —
|
|
52
|
+
these are owned by GitHub Actions automation and the PM. The agent is responsible
|
|
53
|
+
for applying `stage:brainstorming`, `stage:detailing`, and `stage:approval` as
|
|
54
|
+
part of the prescribed workflow above.
|
|
8
55
|
|
|
9
|
-
|
|
56
|
+
Each stage transition requires a fresh explicit signal from the user in the same
|
|
57
|
+
session where the transition happens. These rules have no exceptions.
|
package/SETUP.md
CHANGED
|
@@ -12,6 +12,8 @@ Estimated time: 2-3 hours (including GitHub Projects setup).
|
|
|
12
12
|
- GitHub Projects enabled on your account/organization
|
|
13
13
|
- Implementation agent connected to the repository (e.g., Jules: github.com/google-labs/jules)
|
|
14
14
|
|
|
15
|
+
> **Jules users:** After installing the Jules GitHub App at [github.com/settings/installations](https://github.com/settings/installations), set `JULES_ENABLED=true` at **repo → Settings → Variables → Actions** to enable Jules dispatch. If this variable is absent, agent comments are silently skipped — no false "dispatched" messages.
|
|
16
|
+
|
|
15
17
|
---
|
|
16
18
|
|
|
17
19
|
## Step 1 — Create the GitHub Project Board
|
|
@@ -48,6 +48,10 @@ If any of these files are missing, you are in **Setup Mode**. Do not proceed wit
|
|
|
48
48
|
- a) **No** — *No autonomous implementation agent.*
|
|
49
49
|
- b) **@google-labs-jules** — *Jules (recommended if you don't have one).*
|
|
50
50
|
- c) **Other** — *Enter the agent's handle.*
|
|
51
|
+
|
|
52
|
+
When writing `agent-trigger.yml`, set `{{IMPLEMENTATION_AGENT_LABEL}}` as follows:
|
|
53
|
+
- Jules (`@google-labs-jules`): use `jules` — this is the native label the Jules GitHub App watches. **Do NOT use `agent:jules`** — Jules does not watch that label and the trigger will silently fail.
|
|
54
|
+
- Other agents: use the handle without `@`, lowercase (e.g. `@my-agent` → `my-agent`).
|
|
51
55
|
5. Generate and write the missing files replacing the `{{SCREAMING_SNAKE_CASE}}` placeholders using the templates in `.agentic-pdlc/templates/`.
|
|
52
56
|
6. Offer to run the `gh` commands for labels (`spec:approved`, `pr:in-review`, `pr:approved`, `architecture-violation`).
|
|
53
57
|
7. **`PROJECT_PAT` secret (required for board automation):**
|
|
@@ -98,9 +102,9 @@ If `AGENTS.md` and `docs/pdlc.md` are present, you are in **Execution Mode**.
|
|
|
98
102
|
|
|
99
103
|
### 0. [FIRST] Issue Type Identification
|
|
100
104
|
|
|
101
|
-
**Run before anything else — before
|
|
105
|
+
**Run before anything else — before reading code.**
|
|
102
106
|
|
|
103
|
-
Reading the issue title and body for type inference is exempt from the
|
|
107
|
+
Reading the issue title and body for type inference is exempt from the initial label requirement: it is metadata already present in the request, not code reading or skill invocation.
|
|
104
108
|
|
|
105
109
|
1. Check if issue already has a `type:*` label (`type:us`, `type:task`, `type:bug`, `type:spike`) → if yes, skip to Section 0.1.
|
|
106
110
|
2. Read issue title + body (metadata only — no code reading at this step).
|
|
@@ -116,10 +120,10 @@ Reading the issue title and body for type inference is exempt from the `stage:ex
|
|
|
116
120
|
|
|
117
121
|
| Type | Flow |
|
|
118
122
|
|---|---|
|
|
119
|
-
| `type:us` |
|
|
120
|
-
| `type:task` |
|
|
121
|
-
| `type:bug` |
|
|
122
|
-
| `type:spike` |
|
|
123
|
+
| `type:us` | brainstorming → Gate 1 → detailing → approval |
|
|
124
|
+
| `type:task` | brainstorming → Gate 1 → detailing → approval |
|
|
125
|
+
| `type:bug` | brainstorming → Gate 1 → detailing → approval |
|
|
126
|
+
| `type:spike` | brainstorming → Gate 1 → detailing → conclusion comment (never reaches Development) |
|
|
123
127
|
|
|
124
128
|
### 0.1 Board Labels — Mandatory at Every State Transition
|
|
125
129
|
|
|
@@ -127,11 +131,10 @@ These label commands are non-negotiable. They run **before** the activity they a
|
|
|
127
131
|
|
|
128
132
|
| When | Command |
|
|
129
133
|
|---|---|
|
|
130
|
-
| Before reading any code / invoking any skill | `gh issue edit <N> --add-label "stage:
|
|
131
|
-
| Before presenting architecture approaches | `gh issue edit <N> --add-label "stage:brainstorming" --remove-label "stage:exploration"` |
|
|
134
|
+
| Before reading any code / invoking any skill | `gh issue edit <N> --add-label "stage:brainstorming"` |
|
|
132
135
|
| Before writing the technical spec | `gh issue edit <N> --add-label "stage:detailing" --remove-label "stage:brainstorming"` |
|
|
133
136
|
|
|
134
|
-
No investigation, no skill invocation, no code reading happens before `stage:
|
|
137
|
+
No investigation, no skill invocation, no code reading happens before `stage:brainstorming` is applied. No spec writing starts before `stage:detailing` is set (and `stage:brainstorming` removed).
|
|
135
138
|
|
|
136
139
|
### 0.1 PR Stage Gate — Non-Negotiable
|
|
137
140
|
|
|
@@ -147,7 +150,7 @@ git checkout -b hotfix/<N>-<description>
|
|
|
147
150
|
```
|
|
148
151
|
|
|
149
152
|
### 1. Daily Upstream Loop
|
|
150
|
-
Your job is to move issues from "💡 Idea
|
|
153
|
+
Your job is to move issues from "💡 Idea" to "📐 Detail Solution".
|
|
151
154
|
When asked to work on a feature, you will:
|
|
152
155
|
- Explore the code context.
|
|
153
156
|
- Present architectural approaches (Brainstorming).
|
|
@@ -26,8 +26,8 @@ fi
|
|
|
26
26
|
|
|
27
27
|
LABELS=$(gh issue view "$ISSUE_NUM" --json labels --jq '[.labels[].name] | join(" ")' 2>/dev/null || echo "")
|
|
28
28
|
|
|
29
|
-
if echo "$LABELS" | grep -qw "
|
|
30
|
-
echo "✅ PDLC: Issue #$ISSUE_NUM approved — gate passed."
|
|
29
|
+
if echo "$LABELS" | grep -qw "spec:approved"; then
|
|
30
|
+
echo "✅ PDLC: Issue #$ISSUE_NUM spec approved by PM — gate passed."
|
|
31
31
|
exit 0
|
|
32
32
|
fi
|
|
33
33
|
|
|
@@ -39,6 +39,6 @@ fi
|
|
|
39
39
|
STAGE=$(echo "$LABELS" | tr ' ' '\n' | grep "^stage:" | head -1 || echo "none")
|
|
40
40
|
echo "❌ PDLC Stage Gate: Issue #$ISSUE_NUM is not approved."
|
|
41
41
|
echo " Current stage: $STAGE"
|
|
42
|
-
echo " Required:
|
|
42
|
+
echo " Required: spec:approved (set by human PM) or stage:development label on the issue."
|
|
43
43
|
echo " Emergency bypass: rename branch to hotfix/<issue-number>-<description>."
|
|
44
44
|
exit 1
|
package/bin/cli.js
CHANGED
|
@@ -227,6 +227,16 @@ async function runSetup() {
|
|
|
227
227
|
}
|
|
228
228
|
}
|
|
229
229
|
|
|
230
|
+
const defaultLabels = [
|
|
231
|
+
'bug', 'documentation', 'duplicate', 'enhancement',
|
|
232
|
+
'good first issue', 'help wanted', 'invalid', 'question', 'wontfix'
|
|
233
|
+
];
|
|
234
|
+
for (const label of defaultLabels) {
|
|
235
|
+
try {
|
|
236
|
+
execFileSync('gh', ['label', 'delete', label, '--repo', repo, '--yes'], { stdio: 'ignore' });
|
|
237
|
+
} catch (_) {}
|
|
238
|
+
}
|
|
239
|
+
|
|
230
240
|
// Project V2
|
|
231
241
|
console.log(`\n${cyan}${i18n.creating_project}${reset}`);
|
|
232
242
|
let ownerId, projectId, projectNumber;
|
|
@@ -240,13 +250,13 @@ async function runSetup() {
|
|
|
240
250
|
}
|
|
241
251
|
|
|
242
252
|
const projectCreateRaw = execFileSync('gh', ['api', 'graphql', '-f', 'query=mutation($owner: ID!, $title: String!) { createProjectV2(input: {ownerId: $owner, title: $title}) { projectV2 { id number } } }', '-f', `owner=${ownerId}`, '-f', `title=${boardName}`], { stdio: ['ignore', 'pipe', 'pipe'] }).toString().trim();
|
|
243
|
-
const projectCreateResponse = JSON.parse(projectCreateRaw);
|
|
244
|
-
if (projectCreateResponse
|
|
253
|
+
const projectCreateResponse = projectCreateRaw ? JSON.parse(projectCreateRaw) : null;
|
|
254
|
+
if (projectCreateResponse?.errors) {
|
|
245
255
|
throw new Error(projectCreateResponse.errors.map(e => e.message).join('; '));
|
|
246
256
|
}
|
|
247
|
-
const projectCreateData = projectCreateResponse
|
|
248
|
-
projectId = projectCreateData
|
|
249
|
-
projectNumber = projectCreateData
|
|
257
|
+
const projectCreateData = projectCreateResponse?.data?.createProjectV2?.projectV2;
|
|
258
|
+
projectId = projectCreateData?.id;
|
|
259
|
+
projectNumber = projectCreateData?.number;
|
|
250
260
|
|
|
251
261
|
console.log(` ${i18n.project_ok}${projectId})`);
|
|
252
262
|
|
|
@@ -280,7 +290,7 @@ async function runSetup() {
|
|
|
280
290
|
|
|
281
291
|
if (statusFieldId) {
|
|
282
292
|
const columns = [
|
|
283
|
-
{ name: "💡 Idea", description: "
|
|
293
|
+
{ name: "💡 Idea - No move to Exploration directly", description: "Just tell your agent to work on issue #XX", color: "GRAY" },
|
|
284
294
|
{ name: "🔍 Exploration", description: "AI is analyzing code and context", color: "PURPLE" },
|
|
285
295
|
{ name: "🧠 Brainstorming", description: "AI proposed approaches and trade-offs", color: "PINK" },
|
|
286
296
|
{ name: "📐 Detail Solution", description: "AI is writing the technical spec", color: "BLUE" },
|
|
@@ -314,7 +324,8 @@ async function runSetup() {
|
|
|
314
324
|
|
|
315
325
|
const updateOutput = execFileSync('gh', ['api', 'graphql', '--input', '-'], { input: queryPayload }).toString().trim();
|
|
316
326
|
const jsonResponse = updateOutput ? JSON.parse(updateOutput) : null;
|
|
317
|
-
const returnedOptions = jsonResponse?.data?.updateProjectV2Field?.projectV2Field?.options ||
|
|
327
|
+
const returnedOptions = jsonResponse?.data?.updateProjectV2Field?.projectV2Field?.options ||
|
|
328
|
+
jsonResponse?.data?.updateProjectV2SingleSelectField?.projectV2SingleSelectField?.options || [];
|
|
318
329
|
|
|
319
330
|
for (const opt of returnedOptions) {
|
|
320
331
|
optionMap[opt.name] = opt.id;
|
package/docs/flow.md
CHANGED
|
@@ -5,8 +5,7 @@ This document describes the full lifecycle of a card on the agentic-pdlc board
|
|
|
5
5
|
```mermaid
|
|
6
6
|
stateDiagram-v2
|
|
7
7
|
direction LR
|
|
8
|
-
Idea -->
|
|
9
|
-
Exploration --> Brainstorming : stage:brainstorming
|
|
8
|
+
Idea --> Brainstorming : stage:brainstorming
|
|
10
9
|
Brainstorming --> Detailing : Gate 1 (PM)
|
|
11
10
|
Detailing --> Approval : stage:approval
|
|
12
11
|
Approval --> Development : Gate 2 (spec:approved)
|
|
@@ -23,7 +22,7 @@ stateDiagram-v2
|
|
|
23
22
|
|------|---------------|
|
|
24
23
|
| **PM** (human) | Selects issues for the sprint, approves brainstorm (Gate 1), approves spec (Gate 2) |
|
|
25
24
|
| **TL / Reviewer** (human) | Co-approves spec for technical decisions (Gate 2), reviews and approves the PR (Gate 3) |
|
|
26
|
-
| **Claude** |
|
|
25
|
+
| **Claude** | Brainstorming (context reading + approaches), spec writing — acts on PM instruction via chat or via upstream label |
|
|
27
26
|
| **Implementation Agent** | Implementation, running tests, opening the PR |
|
|
28
27
|
| **Workflow** | All label swaps and card movements — the source of truth for board state |
|
|
29
28
|
|
|
@@ -38,33 +37,22 @@ Issue exists in the backlog with no `stage:` label.
|
|
|
38
37
|
|
|
39
38
|
---
|
|
40
39
|
|
|
41
|
-
### 🔍 Exploration
|
|
42
|
-
|
|
43
|
-
**Trigger (two equivalent paths):**
|
|
44
|
-
- PM tells Claude directly in chat → Claude adds `stage:exploration` to the issue, OR
|
|
45
|
-
- PM adds `stage:exploration` directly to the issue
|
|
46
|
-
|
|
47
|
-
**Workflow:** `project-automation.yml` detects `stage:exploration` → moves card to Exploration.
|
|
48
|
-
|
|
49
|
-
**Who works:** Claude reads relevant code and context.
|
|
50
|
-
|
|
51
|
-
---
|
|
52
|
-
|
|
53
40
|
### 🧠 Brainstorming
|
|
54
41
|
|
|
55
|
-
**Trigger:** Claude
|
|
42
|
+
**Trigger:** PM tells Claude to work on the issue (in chat) or applies `stage:brainstorming` directly.
|
|
56
43
|
|
|
57
44
|
**Actions:**
|
|
58
|
-
- Claude
|
|
59
|
-
- Claude
|
|
45
|
+
- Claude applies `stage:brainstorming` immediately
|
|
46
|
+
- Claude reads relevant code and context
|
|
47
|
+
- Claude posts a single comment with: (1) brief problem summary, (2) 2–3 proposed approaches with trade-offs
|
|
60
48
|
|
|
61
|
-
**Workflow:** `project-automation.yml` moves card to Brainstorming.
|
|
49
|
+
**Workflow:** `project-automation.yml` detects `stage:brainstorming` → moves card to Brainstorming.
|
|
62
50
|
|
|
63
51
|
**⏸ Human gate (Gate 1 — PM):** PM reads the brainstorming comment and selects an approach. This can be done:
|
|
64
52
|
- By commenting on the issue (e.g., "Option A", "Go with B", "approved", "lgtm")
|
|
65
53
|
- By telling Claude in chat
|
|
66
54
|
|
|
67
|
-
**Note on implicit approval:**
|
|
55
|
+
**Note on implicit approval:** Selecting one option (e.g., "Option A") counts as implicit approval. The `upstream-gate.yml` detects option-selection patterns in addition to explicit approval words.
|
|
68
56
|
|
|
69
57
|
---
|
|
70
58
|
|
|
@@ -141,7 +129,6 @@ Issue exists in the backlog with no `stage:` label.
|
|
|
141
129
|
|
|
142
130
|
| Label | Added by | Removed by |
|
|
143
131
|
|-------|----------|------------|
|
|
144
|
-
| `stage:exploration` | PM (human) or Claude | Claude |
|
|
145
132
|
| `stage:brainstorming` | Claude | `upstream-gate.yml` |
|
|
146
133
|
| `stage:detailing` | `upstream-gate.yml` | Claude |
|
|
147
134
|
| `stage:approval` | Claude | `agent-trigger.yml` |
|
package/docs/pdlc.md
CHANGED
|
@@ -4,9 +4,8 @@
|
|
|
4
4
|
|
|
5
5
|
| Column | Meaning | Who moves the card |
|
|
6
6
|
|---|---|---|
|
|
7
|
-
| 💡 Idea
|
|
8
|
-
|
|
|
9
|
-
| 🧠 Brainstorming | AI proposed approaches and trade-offs | Label `stage:brainstorming` |
|
|
7
|
+
| 💡 Idea | Backlog — tell agent: "work on issue #XX" | Don't move manually |
|
|
8
|
+
| 🧠 Brainstorming | AI reading context, proposing approaches and trade-offs | Label `stage:brainstorming` |
|
|
10
9
|
| 📐 Detail Solution | AI is writing the technical spec | Label `stage:detailing` |
|
|
11
10
|
| ✅ Approval | Your turn, awaiting `spec:approved` label | Label `spec:approved` |
|
|
12
11
|
| ⚙️ Development | AI implementing the spec | Label `stage:development` |
|
|
@@ -32,7 +31,6 @@ REPO = {{REPO_OWNER}}/{{REPO_NAME}}
|
|
|
32
31
|
| Column | Option ID |
|
|
33
32
|
|---|---|
|
|
34
33
|
| 💡 Idea | `{{ID_IDEA}}` |
|
|
35
|
-
| 🔍 Exploration | `{{ID_EXPLORATION}}` |
|
|
36
34
|
| 🧠 Brainstorming | `{{ID_BRAINSTORMING}}` |
|
|
37
35
|
| 📐 Detail Solution | `{{ID_DETAIL}}` |
|
|
38
36
|
| ✅ Approval | `{{ID_APPROVAL}}` |
|
|
@@ -65,7 +63,6 @@ REPO = {{REPO_OWNER}}/{{REPO_NAME}}
|
|
|
65
63
|
|
|
66
64
|
| Label | Entity | Color | Meaning |
|
|
67
65
|
|---|---|---|---|
|
|
68
|
-
| `stage:exploration` | Issue | Purple | Issue is being evaluated |
|
|
69
66
|
| `stage:brainstorming` | Issue | Pink | Proposed approaches awaiting PM gate |
|
|
70
67
|
| `stage:detailing` | Issue | Blue | Technical spec is being written |
|
|
71
68
|
| `stage:development` | Issue | Orange | Agent is implementing the spec |
|
|
@@ -73,15 +70,15 @@ REPO = {{REPO_OWNER}}/{{REPO_NAME}}
|
|
|
73
70
|
| `pr:in-review` | PR | Yellow | Awaiting code review |
|
|
74
71
|
| `pr:approved` | PR | Green | Code review approved |
|
|
75
72
|
| `type:us` | Issue | Blue | New feature or behavioral change — full flow |
|
|
76
|
-
| `type:task` | Issue | Yellow | Operational/non-functional change —
|
|
77
|
-
| `type:bug` | Issue | Red | Something broken —
|
|
73
|
+
| `type:task` | Issue | Yellow | Operational/non-functional change — full flow |
|
|
74
|
+
| `type:bug` | Issue | Red | Something broken — full flow |
|
|
78
75
|
| `type:spike` | Issue | Gray | Research/evaluation — never reaches Development |
|
|
79
76
|
|
|
80
77
|
## Approval Gates
|
|
81
78
|
|
|
82
79
|
**Gate 1 — PM/Ideation (Brainstorming):**
|
|
83
|
-
|
|
84
|
-
Format: *"Approved — proceed with option X."*
|
|
80
|
+
Agent presents problem summary + 2–3 solution options in a single message. You select an approach.
|
|
81
|
+
Format: *"Option X"* or *"Go with B"* or *"Approved — proceed with option X."*
|
|
85
82
|
|
|
86
83
|
**Gate 2 — Tech Lead (Spec):**
|
|
87
84
|
You add the `spec:approved` label to the issue after reviewing the technical spec in the body.
|
|
@@ -93,13 +90,24 @@ The `type:*` label is the authoritative signal — set automatically by the agen
|
|
|
93
90
|
|
|
94
91
|
| Label | Flow |
|
|
95
92
|
|---|---|
|
|
96
|
-
| `type:us` |
|
|
97
|
-
| `type:task` |
|
|
98
|
-
| `type:bug` |
|
|
99
|
-
| `type:spike` |
|
|
93
|
+
| `type:us` | brainstorming → Gate 1 → detailing → approval |
|
|
94
|
+
| `type:task` | brainstorming → Gate 1 → detailing → approval |
|
|
95
|
+
| `type:bug` | brainstorming → Gate 1 → detailing → approval |
|
|
96
|
+
| `type:spike` | brainstorming → Gate 1 → detailing → conclusion comment (never reaches Development) |
|
|
100
97
|
|
|
101
98
|
If no `type:*` label present and agent confidence < 85%, defaults to `type:us` (safe fallback — never skips gates by omission).
|
|
102
99
|
|
|
100
|
+
## Bypass Mechanism
|
|
101
|
+
|
|
102
|
+
Agents MUST NOT skip any stage. The ONLY authorized bypasses are:
|
|
103
|
+
|
|
104
|
+
| Mechanism | Who authorizes | What it bypasses |
|
|
105
|
+
|---|---|---|
|
|
106
|
+
| `human-approved` label on issue | PM (human) only | All stage gates |
|
|
107
|
+
| Branch prefix `hotfix/` | PM (human) only | PR gate only |
|
|
108
|
+
|
|
109
|
+
Agents MUST NOT self-authorize a bypass. Stop and ask the PM explicitly.
|
|
110
|
+
|
|
103
111
|
## Definition of Done
|
|
104
112
|
|
|
105
113
|
An issue is truly done when:
|
package/package.json
CHANGED
|
@@ -75,7 +75,7 @@ jobs:
|
|
|
75
75
|
console.log(`Issue #${number} → Development`);
|
|
76
76
|
|
|
77
77
|
- name: Comment on issue to trigger agent and prevent race conditions
|
|
78
|
-
if: ${{ !contains('{{IMPLEMENTATION_AGENT_LABEL}}', '{{') }}
|
|
78
|
+
if: ${{ !contains('{{IMPLEMENTATION_AGENT_LABEL}}', '{{') && vars.JULES_ENABLED == 'true' }}
|
|
79
79
|
uses: actions/github-script@v7
|
|
80
80
|
with:
|
|
81
81
|
github-token: ${{ secrets.GITHUB_TOKEN }}
|
|
@@ -118,7 +118,7 @@ jobs:
|
|
|
118
118
|
contents: read
|
|
119
119
|
steps:
|
|
120
120
|
- name: Comment on issue to trigger agent
|
|
121
|
-
if: ${{ !contains('{{IMPLEMENTATION_AGENT_LABEL}}', '{{') }}
|
|
121
|
+
if: ${{ !contains('{{IMPLEMENTATION_AGENT_LABEL}}', '{{') && vars.JULES_ENABLED == 'true' }}
|
|
122
122
|
uses: actions/github-script@v7
|
|
123
123
|
with:
|
|
124
124
|
github-token: ${{ secrets.GITHUB_TOKEN }}
|
|
@@ -38,3 +38,17 @@ jobs:
|
|
|
38
38
|
if: ${{ !contains('{{TEST_COMMAND}}', '{{') }}
|
|
39
39
|
run: |
|
|
40
40
|
{{TEST_COMMAND}}
|
|
41
|
+
|
|
42
|
+
ci-status:
|
|
43
|
+
name: Sentinel / CI
|
|
44
|
+
needs: [validate]
|
|
45
|
+
if: always()
|
|
46
|
+
runs-on: ubuntu-latest
|
|
47
|
+
steps:
|
|
48
|
+
- name: Check validate result
|
|
49
|
+
run: |
|
|
50
|
+
result="${{ needs.validate.result }}"
|
|
51
|
+
if [[ "$result" != "success" && "$result" != "skipped" ]]; then
|
|
52
|
+
echo "CI failed: validate=$result"
|
|
53
|
+
exit 1
|
|
54
|
+
fi
|
|
@@ -8,7 +8,6 @@ on:
|
|
|
8
8
|
env:
|
|
9
9
|
PROJECT_ID: "{{PROJECT_ID}}"
|
|
10
10
|
STATUS_FIELD_ID: "{{STATUS_FIELD_ID}}"
|
|
11
|
-
STATUS_EXPLORATION: "{{ID_EXPLORATION}}"
|
|
12
11
|
STATUS_BRAINSTORMING: "{{ID_BRAINSTORMING}}"
|
|
13
12
|
STATUS_DETAILING: "{{ID_DETAILING}}"
|
|
14
13
|
STATUS_APPROVAL: "{{ID_APPROVAL}}"
|
|
@@ -35,7 +34,6 @@ jobs:
|
|
|
35
34
|
const projectId = process.env.PROJECT_ID;
|
|
36
35
|
const statusFieldId = process.env.STATUS_FIELD_ID;
|
|
37
36
|
const envVars = {
|
|
38
|
-
'STATUS_EXPLORATION': process.env.STATUS_EXPLORATION,
|
|
39
37
|
'STATUS_BRAINSTORMING': process.env.STATUS_BRAINSTORMING,
|
|
40
38
|
'STATUS_DETAILING': process.env.STATUS_DETAILING,
|
|
41
39
|
'STATUS_APPROVAL': process.env.STATUS_APPROVAL,
|
|
@@ -6,13 +6,12 @@ on:
|
|
|
6
6
|
pull_request_review:
|
|
7
7
|
types: [submitted]
|
|
8
8
|
issues:
|
|
9
|
-
types: [labeled, edited]
|
|
9
|
+
types: [labeled, edited, closed]
|
|
10
10
|
|
|
11
11
|
env:
|
|
12
12
|
PROJECT_ID: "{{PROJECT_ID}}"
|
|
13
13
|
STATUS_FIELD_ID: "{{STATUS_FIELD_ID}}"
|
|
14
14
|
STATUS_IDEA: "{{ID_IDEA}}"
|
|
15
|
-
STATUS_EXPLORATION: "{{ID_EXPLORATION}}"
|
|
16
15
|
STATUS_BRAINSTORMING: "{{ID_BRAINSTORMING}}"
|
|
17
16
|
STATUS_DETAILING: "{{ID_DETAILING}}"
|
|
18
17
|
STATUS_APPROVAL: "{{ID_APPROVAL}}"
|
|
@@ -40,10 +39,7 @@ jobs:
|
|
|
40
39
|
let targetStatusId = null;
|
|
41
40
|
let stageName = null;
|
|
42
41
|
|
|
43
|
-
if (labelName === 'stage:
|
|
44
|
-
targetStatusId = process.env.STATUS_EXPLORATION;
|
|
45
|
-
stageName = 'Exploration';
|
|
46
|
-
} else if (labelName === 'stage:brainstorming') {
|
|
42
|
+
if (labelName === 'stage:brainstorming') {
|
|
47
43
|
targetStatusId = process.env.STATUS_BRAINSTORMING;
|
|
48
44
|
stageName = 'Brainstorming';
|
|
49
45
|
} else if (labelName === 'stage:detailing') {
|
|
@@ -302,3 +298,25 @@ jobs:
|
|
|
302
298
|
} else {
|
|
303
299
|
await moveItem(pr.node_id);
|
|
304
300
|
}
|
|
301
|
+
|
|
302
|
+
cleanup-labels-on-close:
|
|
303
|
+
name: Issue closed → strip stage/agent labels
|
|
304
|
+
if: github.event_name == 'issues' && github.event.action == 'closed'
|
|
305
|
+
runs-on: ubuntu-latest
|
|
306
|
+
steps:
|
|
307
|
+
- name: Remove transient labels
|
|
308
|
+
uses: actions/github-script@v7
|
|
309
|
+
with:
|
|
310
|
+
github-token: ${{ secrets.GITHUB_TOKEN }}
|
|
311
|
+
script: |
|
|
312
|
+
const { owner, repo } = context.repo;
|
|
313
|
+
const issue_number = context.payload.issue.number;
|
|
314
|
+
const toRemove = [
|
|
315
|
+
'stage:brainstorming', 'stage:detailing',
|
|
316
|
+
'stage:approval', 'stage:development', 'stage:testing',
|
|
317
|
+
'agent:working', 'qa:needs-work', 'pr:in-review', 'jules'
|
|
318
|
+
];
|
|
319
|
+
for (const label of toRemove) {
|
|
320
|
+
await github.rest.issues.removeLabel({ owner, repo, issue_number, name: label }).catch(() => {});
|
|
321
|
+
}
|
|
322
|
+
console.log(`Issue #${issue_number} labels cleaned up`);
|
|
@@ -13,6 +13,8 @@ jobs:
|
|
|
13
13
|
qa:
|
|
14
14
|
name: AC Coverage Verification (GitHub Models)
|
|
15
15
|
runs-on: ubuntu-latest
|
|
16
|
+
env:
|
|
17
|
+
PROJECT_PAT: ${{ secrets.PROJECT_PAT }}
|
|
16
18
|
steps:
|
|
17
19
|
- uses: actions/checkout@v4
|
|
18
20
|
with:
|
|
@@ -59,7 +61,7 @@ jobs:
|
|
|
59
61
|
--max-time 30 || echo "API_ERROR")
|
|
60
62
|
|
|
61
63
|
if [ "$RESPONSE" = "API_ERROR" ]; then
|
|
62
|
-
gh
|
|
64
|
+
GH_TOKEN="$PROJECT_PAT" gh api "repos/${GITHUB_REPOSITORY}/issues/${PR_NUMBER}/labels" --method POST -f 'labels[]=infra:qa-broken'
|
|
63
65
|
gh pr comment "$PR_NUMBER" --body "🤖 **QA Agent:** Could not reach GitHub Models API. Manual review required."
|
|
64
66
|
exit 0
|
|
65
67
|
fi
|
|
@@ -68,12 +70,12 @@ jobs:
|
|
|
68
70
|
EXPLANATION=$(echo "$RESPONSE" | python3 -c 'import json,sys; d=json.load(sys.stdin); t=d.get("choices",[{}])[0].get("message",{}).get("content","").strip(); lines=t.split("\n",1); print(lines[1].strip() if len(lines)>1 else "")')
|
|
69
71
|
|
|
70
72
|
if echo "$VERDICT" | grep -q "^PASS"; then
|
|
71
|
-
gh
|
|
73
|
+
GH_TOKEN="$PROJECT_PAT" gh api "repos/${GITHUB_REPOSITORY}/issues/${PR_NUMBER}/labels" --method POST -f 'labels[]=qa:approved'
|
|
72
74
|
gh pr comment "$PR_NUMBER" --body "🤖 **QA Agent:** AC coverage verified. ${EXPLANATION}"
|
|
73
75
|
elif echo "$VERDICT" | grep -q "^FAIL"; then
|
|
74
|
-
gh
|
|
76
|
+
GH_TOKEN="$PROJECT_PAT" gh api "repos/${GITHUB_REPOSITORY}/issues/${PR_NUMBER}/labels" --method POST -f 'labels[]=qa:needs-work'
|
|
75
77
|
gh pr comment "$PR_NUMBER" --body "🤖 **QA Agent:** AC coverage insufficient. ${EXPLANATION}"
|
|
76
78
|
else
|
|
77
|
-
gh
|
|
79
|
+
GH_TOKEN="$PROJECT_PAT" gh api "repos/${GITHUB_REPOSITORY}/issues/${PR_NUMBER}/labels" --method POST -f 'labels[]=infra:qa-broken'
|
|
78
80
|
gh pr comment "$PR_NUMBER" --body "🤖 **QA Agent:** Could not parse GitHub Models response. Manual review required."
|
|
79
81
|
fi
|
package/templates/AGENTS.md
CHANGED
|
@@ -30,7 +30,7 @@ Always start from the current `main` HEAD. Never work over stale snapshots.
|
|
|
30
30
|
## Mandatory Workflow
|
|
31
31
|
|
|
32
32
|
0. **Identity**: Always prefix your GitHub comments with `🤖 **Agent:** ` to distinguish yourself.
|
|
33
|
-
1. **Initial State**: When beginning work on a new issue, your very first action must be to apply the `stage:
|
|
33
|
+
1. **Initial State**: When beginning work on a new issue, your very first action must be to apply the `stage:brainstorming` label using the GitHub CLI (`gh issue edit <N> --add-label "stage:brainstorming"`).
|
|
34
34
|
2. Read the issue entirely — understand its type (US/BUG/TASK/SPIKE) and the Acceptance Criteria.
|
|
35
35
|
3. Read `docs/pdlc.md` — understand the PDLC and the Definition of Done in this project.
|
|
36
36
|
4. Read all files mentioned in the issue's technical context.
|
|
@@ -64,6 +64,25 @@ When detailing a solution in an issue body, you must **always** include both the
|
|
|
64
64
|
- `path/to/file.ts` — what changes
|
|
65
65
|
```
|
|
66
66
|
|
|
67
|
+
## Stage Transition Rules (non-negotiable)
|
|
68
|
+
|
|
69
|
+
MUST apply `stage:brainstorming` label immediately on starting work — before reading
|
|
70
|
+
any code, before invoking any skill. Then read context and present problem summary
|
|
71
|
+
+ 2–3 solution options in a single message.
|
|
72
|
+
|
|
73
|
+
MUST NOT add `stage:detailing` label until the user has explicitly selected
|
|
74
|
+
an approach in the current conversation turn. Work done in a prior
|
|
75
|
+
planning session does NOT count as confirmation.
|
|
76
|
+
|
|
77
|
+
MUST NOT add `spec:approved`, `stage:development`, or manually add
|
|
78
|
+
`stage:approval` — these represent final human approval or the result of it.
|
|
79
|
+
`stage:approval` is only set by system automation after you provide a complete
|
|
80
|
+
spec for human review. Adding them manually triggers irreversible automation
|
|
81
|
+
(Jules dispatch, board move).
|
|
82
|
+
|
|
83
|
+
Each stage transition requires a fresh explicit signal from the user in the same
|
|
84
|
+
session where the transition happens. These rules have no exceptions.
|
|
85
|
+
|
|
67
86
|
## Pipeline Updates
|
|
68
87
|
|
|
69
88
|
To add or configure optional agents (Jules, QA Agent, Sentinel) at any time:
|
|
@@ -80,7 +99,11 @@ Run this when the user says anything like "update the pipeline", "update the boa
|
|
|
80
99
|
- Never open a PR without passing the tests.
|
|
81
100
|
- Never implement beyond the immediate scope of the issue.
|
|
82
101
|
- Never create future-proofing abstractions for hypothetical features.
|
|
83
|
-
-
|
|
102
|
+
- The agent MUST NOT apply these labels under any circumstances (PM only):
|
|
103
|
+
- `spec:approved`: triggers Jules dispatch + board move to Development.
|
|
104
|
+
- `qa:approved`: triggers board move to Code Review.
|
|
105
|
+
- `qa:needs-work`: signals the PR requires changes and halts the flow.
|
|
106
|
+
- Never add or remove stage:* labels manually, except for stage:brainstorming as the initial label when starting work. All other stage transitions are owned by GitHub Actions automation and the PM.
|
|
84
107
|
{{EXTRA_DONT}}
|
|
85
108
|
|
|
86
109
|
## Project Standards
|
package/templates/docs/pdlc.md
CHANGED
|
@@ -4,9 +4,8 @@
|
|
|
4
4
|
|
|
5
5
|
| Column | Meaning | Who moves the card |
|
|
6
6
|
|---|---|---|
|
|
7
|
-
| 💡 Idea
|
|
8
|
-
|
|
|
9
|
-
| 🧠 Brainstorming | Claude proposed approaches, awaiting PM gate | Label `stage:brainstorming` |
|
|
7
|
+
| 💡 Idea | Backlog — tell agent: "work on issue #XX" | Don't move manually |
|
|
8
|
+
| 🧠 Brainstorming | AI reading context, proposing approaches and trade-offs | Label `stage:brainstorming` |
|
|
10
9
|
| 📐 Detail Solution | Claude is writing the technical spec | Label `stage:detailing` |
|
|
11
10
|
| ✅ Approval | Spec ready, awaiting `spec:approved` label | Label `spec:approved` |
|
|
12
11
|
| ⚙️ Development | Agent implementing the spec | Label `stage:development` |
|
|
@@ -37,7 +36,6 @@ REPO = {{REPO_OWNER}}/{{REPO_NAME}}
|
|
|
37
36
|
| Column | Option ID |
|
|
38
37
|
|---|---|
|
|
39
38
|
| 💡 Idea | `{{ID_IDEA}}` |
|
|
40
|
-
| 🔍 Exploration | `{{ID_EXPLORATION}}` |
|
|
41
39
|
| 🧠 Brainstorming | `{{ID_BRAINSTORMING}}` |
|
|
42
40
|
| 📐 Detail Solution | `{{ID_DETAIL}}` |
|
|
43
41
|
| ✅ Approval | `{{ID_APPROVAL}}` |
|
|
@@ -70,7 +68,6 @@ REPO = {{REPO_OWNER}}/{{REPO_NAME}}
|
|
|
70
68
|
|
|
71
69
|
| Label | Entity | Color | Meaning |
|
|
72
70
|
|---|---|---|---|
|
|
73
|
-
| `stage:exploration` | Issue | Purple | Issue is being evaluated |
|
|
74
71
|
| `stage:brainstorming` | Issue | Pink | Proposed approaches awaiting PM gate |
|
|
75
72
|
| `stage:detailing` | Issue | Blue | Technical spec is being written |
|
|
76
73
|
| `stage:development` | Issue | Orange | Agent is implementing the spec |
|
|
@@ -81,15 +78,15 @@ REPO = {{REPO_OWNER}}/{{REPO_NAME}}
|
|
|
81
78
|
| `qa:needs-work` | PR | Red | QA Agent failed — PR needs changes |
|
|
82
79
|
| `infra:qa-broken` | PR | Orange | QA Agent error — manual review required |
|
|
83
80
|
| `type:us` | Issue | Blue | New feature or behavioral change — full flow |
|
|
84
|
-
| `type:task` | Issue | Yellow | Operational/non-functional change —
|
|
85
|
-
| `type:bug` | Issue | Red | Something broken —
|
|
81
|
+
| `type:task` | Issue | Yellow | Operational/non-functional change — full flow |
|
|
82
|
+
| `type:bug` | Issue | Red | Something broken — full flow |
|
|
86
83
|
| `type:spike` | Issue | Gray | Research/evaluation — never reaches Development |
|
|
87
84
|
|
|
88
85
|
## Approval Gates
|
|
89
86
|
|
|
90
87
|
**Gate 1 — PM/Ideation (Brainstorming):**
|
|
91
|
-
|
|
92
|
-
Format: *"Approved — proceed with option X."*
|
|
88
|
+
Agent presents problem summary + 2–3 solution options in a single message. You select an approach.
|
|
89
|
+
Format: *"Option X"* or *"Go with B"* or *"Approved — proceed with option X."*
|
|
93
90
|
|
|
94
91
|
**Gate 2 — Tech Lead (Spec):**
|
|
95
92
|
You add the `spec:approved` label to the issue after reviewing the technical spec in the body.
|
|
@@ -101,13 +98,24 @@ The `type:*` label is the authoritative signal — set automatically by the agen
|
|
|
101
98
|
|
|
102
99
|
| Label | Flow |
|
|
103
100
|
|---|---|
|
|
104
|
-
| `type:us` |
|
|
105
|
-
| `type:task` |
|
|
106
|
-
| `type:bug` |
|
|
107
|
-
| `type:spike` |
|
|
101
|
+
| `type:us` | brainstorming → Gate 1 → detailing → approval |
|
|
102
|
+
| `type:task` | brainstorming → Gate 1 → detailing → approval |
|
|
103
|
+
| `type:bug` | brainstorming → Gate 1 → detailing → approval |
|
|
104
|
+
| `type:spike` | brainstorming → Gate 1 → detailing → conclusion comment (never reaches Development) |
|
|
108
105
|
|
|
109
106
|
If no `type:*` label present and agent confidence < 85%, defaults to `type:us` (safe fallback — never skips gates by omission).
|
|
110
107
|
|
|
108
|
+
## Bypass Mechanism
|
|
109
|
+
|
|
110
|
+
Agents MUST NOT skip any stage. The ONLY authorized bypasses are:
|
|
111
|
+
|
|
112
|
+
| Mechanism | Who authorizes | What it bypasses |
|
|
113
|
+
|---|---|---|
|
|
114
|
+
| `human-approved` label on issue | PM (human) only | All stage gates |
|
|
115
|
+
| Branch prefix `hotfix/` | PM (human) only | PR gate only |
|
|
116
|
+
|
|
117
|
+
Agents MUST NOT self-authorize a bypass. Stop and ask the PM explicitly.
|
|
118
|
+
|
|
111
119
|
## Definition of Done
|
|
112
120
|
|
|
113
121
|
An issue is truly done when:
|