Skip to content

Commit acf01d0

Browse files
Fix agentic triage comment submission and reject incomplete summaries (#4706)
* Fix agentic triage safe-output submission Allow jq payload preparation, document one-shot JSON submission, and validate triage summaries before safe-output writes. Preserve comment limits, label restrictions, and incomplete-result reporting. Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com> * Address triage validator review feedback Reject partial safe-output batches, preserve valid C# generics while detecting embedded template placeholders, and prohibit comment resubmission after any attempt. Remove redundant payload-encoding coverage now provided by merged triage preparation. Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com> * Reject marker-prefixed triage drafts and document validation intent Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com> --------- Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
1 parent c7ab087 commit acf01d0

6 files changed

Lines changed: 628 additions & 5 deletions

File tree

‎.github/scripts/tests/README.md‎

Lines changed: 32 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -1,5 +1,37 @@
11
# GitHub Actions Script Tests
22

3+
## Agentic triage output tests
4+
5+
`validate-triage-output.test.cjs` uses the built-in Node.js test runner (Node.js
6+
20 or newer). No npm dependencies are needed. Run it from the repository root:
7+
8+
```bash
9+
node --test .github/scripts/tests/validate-triage-output.test.cjs
10+
```
11+
12+
These tests also run in `verify-aw-lock.yml`. They cover the validation gate
13+
before triage comments and labels are published: complete summaries, explicit
14+
no-op/failure results, placeholder comments, unfinished templates, quoted/fenced
15+
examples, duplicate comments, and mixed success/incomplete outcomes.
16+
They also reject partial safe-output batches with recorded errors while
17+
allowing C# generic type syntax in completed summaries. Draft prefixes such as
18+
`TODO:` and `TBD:` are rejected in required fields and list items; ordinary
19+
mentions of those markers within completed prose remain valid.
20+
Coverage also includes multiline template placeholders, label operations queued
21+
before the summary, and contradictory no-op/incomplete results.
22+
Comparison operators in completed prose remain valid; angle-bracket detection
23+
targets known template fields and instructions rather than arbitrary bracketed text.
24+
Generic arguments such as `Task<Result>` and `List<Item>`, including nested
25+
types, and XML documentation tags are covered by acceptance regressions.
26+
The summary heading must start at the beginning of the body; leading indentation
27+
cannot turn a code-block heading into an accepted Markdown heading.
28+
Initial and follow-up summaries must queue exactly one Waiting for Author label
29+
operation matching the Environment row; on-demand summaries must leave labels
30+
unchanged. Tests also reject diagnostic placeholders within required fields,
31+
duplicate required rows/sections, and unfinished content in additional prose.
32+
33+
## Shell script tests
34+
335
This directory contains tests for the shell scripts used by the
436
[cherry-pick-hotfix](./../../../.github/workflows/cherry-pick-hotfix.yml),
537
[check-milestone](./../../../.github/workflows/check-milestone.yml) and

0 commit comments

Comments
 (0)