getdevintern/devintern / plugins/devintern/skills/devintern-report
devintern-report
VerifiedDiagnose a DevIntern CLI or worker failure and file a redacted bug report with `devintern report`. Use when DevIntern crashes, a task run fails unexpectedly or completes with incorrect behavior, or the user asks to report a bug.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 5.9 KB |
SKILL.md
View on GitHub---
name: devintern-report
description: Diagnose a DevIntern CLI or worker failure and file a redacted bug report with `devintern report`. Use when DevIntern crashes, a task run fails unexpectedly or completes with incorrect behavior, or the user asks to report a bug.
---
# Report a DevIntern bug
Use `devintern report` to file a defect in DevIntern Code, the worker, a pipeline plugin, or an agent integration. The command collects a redacted diagnostic bundle, shows a preview, and submits only after consent.
Reporting is for product defects. Missing credentials, a tracker outage, an uninstalled agent CLI, a usage limit, or a failed verification check are configuration or user-code issues: fix those first or point to the relevant guide instead of filing a report.
## Confirm the command exists
Check `devintern report --help` before using flags. If the installed CLI does not have the command, say so and suggest upgrading `@getdevintern/code`; do not simulate a report or paste logs into a public issue.
Existing Sentry error tracking captures exceptions automatically. Use a manual report to add expected behavior, reproduction steps, impact, or an incorrect result that produced no exception.
## Diagnose before filing
- Run `devintern --version` and `devintern doctor`.
- For a worker run, find the failing run in the run history instead of describing it from memory. `devintern worker runs --status failed --limit 20 --json` lists recent failures (filter with `--automation <id>`, `--origin <o>`, or `--task-key <k>`; an automation run that recorded `succeeded` can still have a report saying it was blocked). `devintern worker runs show <id> --json` returns its stages, `outcomeReason`, and the agent's report. `devintern worker runs logs <id>` lists its harness transcripts, and `devintern worker logs --level warn` shows recent worker warnings and errors. These commands are read-only and need no dashboard, running worker, or network. If `devintern worker --help` does not list them, the installed CLI predates them: use the dashboard instead and suggest upgrading `@getdevintern/code`.
- Read the failing output and the relevant docs page. For tracker, agent, relay, or workspace problems, check [configuration](https://devintern.com/docs/code/configuration), [worker](https://devintern.com/docs/code/worker), and [relay](https://devintern.com/docs/code/relay).
- Try a minimal, read-only reproduction when possible. Do not re-run a destructive or non-idempotent command just to reproduce.
- If the failure is clearly configuration, help fix it and stop. File a report only when the behavior contradicts the docs or looks like a product crash.
## Collect evidence safely
Capture the failing output to a unique temporary file instead of pasting it into chat:
```bash
devintern PROJ-123 --create-pr 2>&1 | tee /tmp/devintern-report-UNIQUE.log
```
Then attach the relevant run and log:
```bash
devintern report --last --log /tmp/devintern-report-UNIQUE.log
```
Never place `.env` contents, session files, tokens, task descriptions, prompts, diffs, or customer data in the title, description, or a log file you attach. The command redacts common secrets, but it cannot recognize every proprietary value.
Prefer `--last`, `--task <KEY>`, or `--run <id>` (the id from `devintern worker runs`) over describing a run from memory. Add `--share-identifiers` only when the user agrees to include the task key and repository slug; the default is redacted.
## Draft the report
Write a title that names the component and symptom, for example `worker stops after second harness failover` or `devintern report --from fails with 413`.
The description should include:
- The exact command, with secrets removed
- Expected behavior and what happened instead
- Whether it reproduces, and any workaround
- Relevant version, harness, tracker, and OS details (the command adds most of these)
- The smallest useful log excerpt, attached through `--log` rather than pasted
Keep task content out. Refer to a ticket by a redacted placeholder unless the user explicitly approves sharing it.
## Preview and get consent
Prepare the payload once, then review and send exactly that payload:
```bash
devintern report \
--title "worker stops after second harness failover" \
--body-file /tmp/devintern-report-notes.md \
--last --log /tmp/devintern-report-UNIQUE.log \
--save /tmp/devintern-report-payload.json
devintern report --from /tmp/devintern-report-payload.json --print
```
Summarize for the user what will be attached (environment, run summary, log sizes) and what was deliberately omitted. Do not dump the full JSON into the conversation; show the title and description and describe the rest. Ask for approval before submitting.
Signed-in users get account follow-up. Anonymous reports need a contact address:
```bash
devintern report --from /tmp/devintern-report-payload.json --email user@example.com --yes
```
Only use `--yes` after the user approved the preview. For scripts, add `--json`; stdout is then a single result object.
If submission fails, the command saves the report locally and prints a `--from` retry path. Keep that path and tell the user; do not discard it. If the endpoint is unreachable and the user cannot wait, `--save` the payload and hand over the file instead.
## Report the outcome
Return the short report id from the command, what was attached, what was omitted, and whether the report is submitted or saved locally. Do not promise a fix or a timeline.
## Avoid
- Filing reports for missing credentials, usage limits, or user-code verification failures
- Creating a public GitHub issue containing logs, task content, or repository identifiers
- Pasting the full `--print` payload into chat
- Running the command with `--yes` before showing the preview
- Reporting another person's data without explicit permission
- Inventing flags, endpoints, or report ids when the installed CLI lacks the command
For payload details, previews, and retention, see [Reporting bugs](https://devintern.com/docs/code/reporting-bugs).
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| slack-gif-creatoranthropics/skillsKnowledge and utilities for creating animated GIFs optimized for Slack. Provides constraints, validation tools, and animation concepts. Use when users request animated GIFs for Slack like "make me a GIF of X doing Y for Slack." | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |
| ponytailDietrichGebert/ponytailLazy senior dev mode for any coding task: the smallest change that fully solves it, and a reply a busy human understands in one read. | Stars: 196k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-debtDietrichGebert/ponytailHarvest every shortcut: comment into one debt ledger, so deferrals get tracked instead of forgotten. One-shot report. | Stars: 196k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-debtDietrichGebert/ponytailList every `shortcut:` comment (and older `ponytail:` ones) as a debt ledger. One-shot report, changes nothing. Use for "ponytail debt", "what did ponytail defer", "list the shortcuts", /ponytail-debt. | Stars: 196k | Updated | Claude CodeCodex | Verified | Install |