disinto-admin/disinto

Fork 0

fix: reproduce-agent formula — primary goal is reproduction, not root cause #320

New issue

Closed

opened 2026-04-06 19:34:57 +00:00 by dev-bot · 1 comment

dev-bot commented

2026-04-06 19:34:57 +00:00

Collaborator

Problem

The reproduce-agent formula currently treats reproduction and root cause analysis as equal goals. In practice, the agent spends all its turns investigating instead of confirming the bug first. When it hits the turn limit, we do not even know if the bug is real.

Reframe

The reproduce-agent has one primary job and one secondary, minor job:

Primary: Can the bug be reproduced?

This is the exit gate. The agent must answer YES or NO before doing anything else.

Read the issue, understand the claimed behavior
Navigate the app via Playwright MCP, follow the reported steps
Observe: does the symptom match the report?
Take screenshots as evidence
Conclude: reproduced or cannot reproduce

If cannot reproduce → label bug-report + rejected, post findings, done.
If inconclusive (timeout, env issues) → label bug-report + blocked, post what was tried, done.

Secondary (minor): Is the cause obvious?

Only after reproduction is confirmed. Quick check only — do not go deep.

Check container logs (docker compose logs) for stack traces or error messages
Check browser console output captured during reproduction
If the cause jumps out (wrong address, missing config, parse error) → note it

If obvious cause → label bug-report + in-progress, create backlog issue with cause, done.
If not obvious → label bug-report + in-triage, post reproduction evidence + logs examined. Triage-agent takes over.

Exit gates

The formula must enforce this order:

1. Can I reproduce it?  →  NO  → rejected / blocked → EXIT
                        →  YES → continue
2. Is the cause obvious? → YES → in-progress + backlog issue → EXIT
                         → NO  → in-triage → EXIT

The agent should spend at most 60% of its turn budget on step 1, reserving 40% for step 2 if reproduction succeeds. If step 1 uses all turns, that is fine — the answer is blocked.

Label combinations

Outcome	Labels applied	Next actor
Reproduced + obvious cause	`bug-report` + `in-progress`	Dev-agent
Reproduced + cause unclear	`bug-report` + `in-triage`	Triage-agent
Cannot reproduce	`bug-report` + `rejected`	Human review
Inconclusive (timeout/error)	`bug-report` + `blocked`	Gardener retries or human

Files

formulas/reproduce.toml — rewrite with primary/secondary structure and exit gates
docker/reproduce/entrypoint-reproduce.sh — update label logic to use combinations

## Problem The reproduce-agent formula currently treats reproduction and root cause analysis as equal goals. In practice, the agent spends all its turns investigating instead of confirming the bug first. When it hits the turn limit, we do not even know if the bug is real. ## Reframe The reproduce-agent has one primary job and one secondary, minor job: ### Primary: Can the bug be reproduced? This is the exit gate. The agent must answer YES or NO before doing anything else. 1. Read the issue, understand the claimed behavior 2. Navigate the app via Playwright MCP, follow the reported steps 3. Observe: does the symptom match the report? 4. Take screenshots as evidence 5. Conclude: **reproduced** or **cannot reproduce** If cannot reproduce → label `bug-report` + `rejected`, post findings, done. If inconclusive (timeout, env issues) → label `bug-report` + `blocked`, post what was tried, done. ### Secondary (minor): Is the cause obvious? Only after reproduction is confirmed. Quick check only — do not go deep. 1. Check container logs (`docker compose logs`) for stack traces or error messages 2. Check browser console output captured during reproduction 3. If the cause jumps out (wrong address, missing config, parse error) → note it If obvious cause → label `bug-report` + `in-progress`, create backlog issue with cause, done. If not obvious → label `bug-report` + `in-triage`, post reproduction evidence + logs examined. Triage-agent takes over. ### Exit gates The formula must enforce this order: ``` 1. Can I reproduce it? → NO → rejected / blocked → EXIT → YES → continue 2. Is the cause obvious? → YES → in-progress + backlog issue → EXIT → NO → in-triage → EXIT ``` The agent should spend at most 60% of its turn budget on step 1, reserving 40% for step 2 if reproduction succeeds. If step 1 uses all turns, that is fine — the answer is `blocked`. ## Label combinations | Outcome | Labels applied | Next actor | |---------|---------------|------------| | Reproduced + obvious cause | `bug-report` + `in-progress` | Dev-agent | | Reproduced + cause unclear | `bug-report` + `in-triage` | Triage-agent | | Cannot reproduce | `bug-report` + `rejected` | Human review | | Inconclusive (timeout/error) | `bug-report` + `blocked` | Gardener retries or human | ## Files - `formulas/reproduce.toml` — rewrite with primary/secondary structure and exit gates - `docker/reproduce/entrypoint-reproduce.sh` — update label logic to use combinations

dev-bot added the

backlog

label 2026-04-06 19:34:57 +00:00

dev-bot referenced this issue

2026-04-06 20:06:26 +00:00

feat: triage agent — deep root cause analysis for reproduced bugs #258

dev-qwen self-assigned this 2026-04-06 20:52:15 +00:00

dev-qwen added

in-progress

and removed

backlog

labels 2026-04-06 20:52:16 +00:00

dev-bot added

blocked

and removed

in-progress

labels 2026-04-06 20:54:02 +00:00

dev-bot commented

2026-04-06 20:54:02 +00:00

Author

Collaborator

Stale in-progress issue detected

Field	Value
Detection reason	`no_active_session_no_open_pr`
Timestamp	`2026-04-06T20:54:02Z`

Status: This issue was labeled in-progress but no active tmux session exists.
Action required: A maintainer should triage this issue.

### Stale in-progress issue detected | Field | Value | |---|---| | Detection reason | `no_active_session_no_open_pr` | | Timestamp | `2026-04-06T20:54:02Z` | **Status:** This issue was labeled `in-progress` but no active tmux session exists. **Action required:** A maintainer should triage this issue.

dev-bot referenced this issue from a commit

2026-04-06 20:54:20 +00:00

fix: fix: reproduce-agent formula — primary goal is reproduction, not root cause (#320)

dev-qwen referenced this issue from a pull request that will close it,

2026-04-06 20:54:28 +00:00

fix: fix: reproduce-agent formula — primary goal is reproduction, not root cause (#320) #330

dev-bot referenced this issue from a commit

2026-04-06 21:01:39 +00:00

fix: fix: reproduce-agent formula — primary goal is reproduction, not root cause (#320)

dev-qwen closed this issue

2026-04-06 21:07:45 +00:00

dev-qwen removed their assignment 2026-04-06 21:07:45 +00:00