`.werkrevisor.toml` adds `feature_with_tests` and `harmless_only` to `auto_approvable`, now that 0.10.0 runs on mih09.
Details: the PR document under `docs/prs/`.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Inaccurate PR document at docs/prs/2026-10-05-PR#8-werkrevisor-allows-feature-with-tests.md:9-12
The PR document says the config adds the two change kinds to the auto_approvable list, but it does not state that harmless_only also becomes auto-approvable. The Problem section only motivates feature_with_tests ("a feature PR ... with purely additive tests can become GREEN"). A reviewer cannot tell from the document why harmless_only is allowed or what it permits. Add one sentence on the reason for harmless_only.
Usage: model `openrouter/anthropic/claude-sonnet-5.5`, LLM done, calls 1, tokens in 8255 / out 330 (reasoning included in out), duration 7.3 s, cost 0.0198 USD (reported), config: defaults + host + repository (.werkrevisor.toml at base 94f059c).
Reviewed head `99fd34d1f690be18bd367d544e43b25bfecf401c` against base `94f059c1e17ad1f8a8e2f3f2fa6ed41c2653ec6c`, werkrevisor 0.10.0.
<!-- werkrevisor:99fd34d1f690be18bd367d544e43b25bfecf401c -->
<!-- werkrevisor-verdict:YELLOW -->
## 🟡 werkrevisor: YELLOW / HUMAN_REVIEW
Change kind: `mixed`.
### Reasons
- **YELLOW** change kind 'mixed' (other files, docs); AI-approvable are only: docs only, tests only added, docs and tests, refactoring with unchanged tests
- **YELLOW** risky paths changed: .werkrevisor.toml (built in)
- **YELLOW** model reports 1 key issue(s)
### Findings
1. **Inaccurate PR document** at `docs/prs/2026-10-05-PR#8-werkrevisor-allows-feature-with-tests.md:9-12`
The PR document says the config adds the two change kinds to the `auto_approvable` list, but it does not state that `harmless_only` also becomes auto-approvable. The Problem section only motivates `feature_with_tests` ("a feature PR ... with purely additive tests can become GREEN"). A reviewer cannot tell from the document why `harmless_only` is allowed or what it permits. Add one sentence on the reason for `harmless_only`.
<sub>
Usage: model `openrouter/anthropic/claude-sonnet-5.5`, LLM done, calls 1, tokens in 8255 / out 330 (reasoning included in out), duration 7.3 s, cost 0.0198 USD (reported), config: defaults + host + repository (.werkrevisor.toml at base 94f059c).<br>
Reviewed head `99fd34d1f690be18bd367d544e43b25bfecf401c` against base `94f059c1e17ad1f8a8e2f3f2fa6ed41c2653ec6c`, werkrevisor 0.10.0.
</sub>
ai-reviewer
requested review from mi 2026-10-05 17:05:12 +02:00
Line Length Violation at docs/prs/2026-10-05-PR#8-werkrevisor-allows-feature-with-tests.md:9-14
The sentence "Now 0.10.0 runs on mih09, so a feature PR of werkrevisor with purely additive tests can become GREEN." is 111 characters, within the 128 limit, but the line "harmless_only is a PR of files listed in [categories] harmless, e.g. editor settings, which cannot change the gate." exceeds 120 characters and mixes definition and example. More importantly, lines 10-11 split one sentence across two lines after a comma, which violates the AGENTS.md rule of one sentence per line only if read as a wrapped sentence; this is acceptable. The real issue is the explanation of why the change was not made earlier ("because the installed 0.9.0 would have refused the unknown values") narrates history rather than describing the problem. Consider stating only that 0.10.0 is now deployed, so the values are valid.
Usage: model `openrouter/anthropic/claude-sonnet-5.5`, LLM done, calls 1, tokens in 8343 / out 462 (reasoning included in out), duration 8.0 s, cost 0.0213 USD (reported), config: defaults + host + repository (.werkrevisor.toml at base 94f059c).
Reviewed head `8b3f8f716b11769b217bec7538b702c7c9dafbb1` against base `94f059c1e17ad1f8a8e2f3f2fa6ed41c2653ec6c`, werkrevisor 0.10.0.
<!-- werkrevisor:8b3f8f716b11769b217bec7538b702c7c9dafbb1 -->
<!-- werkrevisor-verdict:YELLOW -->
## 🟡 werkrevisor: YELLOW / HUMAN_REVIEW
Change kind: `mixed`.
### Reasons
- **YELLOW** change kind 'mixed' (other files, docs); AI-approvable are only: docs only, tests only added, docs and tests, refactoring with unchanged tests
- **YELLOW** risky paths changed: .werkrevisor.toml (built in)
- **YELLOW** model recommends merge with caution
- **YELLOW** model reports 1 key issue(s)
### Findings
1. **Line Length Violation** at `docs/prs/2026-10-05-PR#8-werkrevisor-allows-feature-with-tests.md:9-14`
The sentence "Now 0.10.0 runs on mih09, so a feature PR of werkrevisor with purely additive tests can become GREEN." is 111 characters, within the 128 limit, but the line "`harmless_only` is a PR of files listed in `[categories] harmless`, e.g. editor settings, which cannot change the gate." exceeds 120 characters and mixes definition and example. More importantly, lines 10-11 split one sentence across two lines after a comma, which violates the AGENTS.md rule of one sentence per line only if read as a wrapped sentence; this is acceptable. The real issue is the explanation of why the change was not made earlier ("because the installed 0.9.0 would have refused the unknown values") narrates history rather than describing the problem. Consider stating only that 0.10.0 is now deployed, so the values are valid.
<sub>
Usage: model `openrouter/anthropic/claude-sonnet-5.5`, LLM done, calls 1, tokens in 8343 / out 462 (reasoning included in out), duration 8.0 s, cost 0.0213 USD (reported), config: defaults + host + repository (.werkrevisor.toml at base 94f059c).<br>
Reviewed head `8b3f8f716b11769b217bec7538b702c7c9dafbb1` against base `94f059c1e17ad1f8a8e2f3f2fa6ed41c2653ec6c`, werkrevisor 0.10.0.
</sub>
mi
merged commit 4cf8b99535 into main2026-10-05 17:28:53 +02:00
mi
deleted branch claude-code/werkrevisor-allows-feature-with-tests2026-10-05 17:28:53 +02:00
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
.werkrevisor.tomladdsfeature_with_testsandharmless_onlytoauto_approvable, now that 0.10.0 runs on mih09.Details: the PR document under
docs/prs/.🤖 Generated with Claude Code
🟡 werkrevisor: YELLOW / HUMAN_REVIEW
Change kind:
mixed.Reasons
Findings
-
Usage: model `openrouter/anthropic/claude-sonnet-5.5`, LLM done, calls 1, tokens in 8255 / out 330 (reasoning included in out), duration 7.3 s, cost 0.0198 USD (reported), config: defaults + host + repository (.werkrevisor.toml at baseInaccurate PR document at
docs/prs/2026-10-05-PR#8-werkrevisor-allows-feature-with-tests.md:9-12The PR document says the config adds the two change kinds to the
auto_approvablelist, but it does not state thatharmless_onlyalso becomes auto-approvable. The Problem section only motivatesfeature_with_tests("a feature PR ... with purely additive tests can become GREEN"). A reviewer cannot tell from the document whyharmless_onlyis allowed or what it permits. Add one sentence on the reason forharmless_only.94f059c).Reviewed head `99fd34d1f690be18bd367d544e43b25bfecf401c` against base `94f059c1e17ad1f8a8e2f3f2fa6ed41c2653ec6c`, werkrevisor 0.10.0.
🟡 werkrevisor: YELLOW / HUMAN_REVIEW
Change kind:
mixed.Reasons
Findings
-
Usage: model `openrouter/anthropic/claude-sonnet-5.5`, LLM done, calls 1, tokens in 8343 / out 462 (reasoning included in out), duration 8.0 s, cost 0.0213 USD (reported), config: defaults + host + repository (.werkrevisor.toml at baseLine Length Violation at
docs/prs/2026-10-05-PR#8-werkrevisor-allows-feature-with-tests.md:9-14The sentence "Now 0.10.0 runs on mih09, so a feature PR of werkrevisor with purely additive tests can become GREEN." is 111 characters, within the 128 limit, but the line "
harmless_onlyis a PR of files listed in[categories] harmless, e.g. editor settings, which cannot change the gate." exceeds 120 characters and mixes definition and example. More importantly, lines 10-11 split one sentence across two lines after a comma, which violates the AGENTS.md rule of one sentence per line only if read as a wrapped sentence; this is acceptable. The real issue is the explanation of why the change was not made earlier ("because the installed 0.9.0 would have refused the unknown values") narrates history rather than describing the problem. Consider stating only that 0.10.0 is now deployed, so the values are valid.94f059c).Reviewed head `8b3f8f716b11769b217bec7538b702c7c9dafbb1` against base `94f059c1e17ad1f8a8e2f3f2fa6ed41c2653ec6c`, werkrevisor 0.10.0.