feat(B): 验收 DSL 编译器 + 骨架熵仪器（IR-0004 AC-8/AC-9 rev6，卡 .github#329/#330） - #100
Merged
Conversation
randypanding
force-pushed
the
quality-instruments-b
branch
from
August 25, 2026 00:51
bd7a8e2 to
7794a20
Compare
|
Warning Review limit reachedNext included review available in 2 minutes. View limit detailsLimit details: You’ve used all 10 included reviews currently available. Enable usage-based reviews in Billing to review now. Otherwise, wait until the next included review is available. Review configuration: ⚙️ Run configurationConfiguration used: Organization UI Review profile: CHILL Plan: Pro Plus Run ID: 📒 Files selected for processing (6)
ℹ️ Recent review info⚙️ Run configurationConfiguration used: Organization UI Review profile: CHILL Plan: Pro Plus Run ID: 📒 Files selected for processing (6)
📝 WalkthroughWalkthroughChanges该 PR 新增验收 DSL 编译与 hash 溯源校验工具、可复用 PR 门禁工作流，以及骨架方差分析工具。新增文档说明接口、职责边界、输出格式和自测命令。 DSL 编译与生成物校验
PR 工作流门禁
骨架方差分析与输出
Suggested labels: ✨ Finishing Touches🧪 Generate unit tests (beta)
Comment |
Comment on lines
+60
to
+69
| - name: Checkout 检测器（CI-Workflows 同源副本，只读） | ||
| if: github.event_name == 'pull_request' && inputs.detector-ref != '' | ||
| uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1 | ||
| with: | ||
| repository: Cloudbird-Software/CI-Workflows | ||
| ref: ${{ inputs.detector-ref }} | ||
| path: detector | ||
| persist-credentials: false | ||
|
|
||
| - name: Setup Python 3.11（stdlib only——零 pip 依赖，无网络需求） |
PR Summary by QodoAdd spec DSL compiler/verify gate and skeleton divergence instrument (IR-0004 AC-8/9)
AI Description
Diagram
High-Level Assessment
Files changed (20)
|
Code Review by Qodo
1. Indented AC lists rejected
|
Comment on lines
+89
to
+93
| for d in $(printf '%s\n' "${CHANGED[@]}" | xargs -n1 dirname | sort -u); do | ||
| SPEC="$d/spec.md" | ||
| if [ ! -f "$SPEC" ]; then | ||
| echo "::error::$SPEC 不存在（$d 有变更但缺 spec.md）" | ||
| FAIL=1; continue |
There was a problem hiding this comment.
1. Nested changes use wrong spec 🐞 Bug ≡ Correctness
The workflow maps each changed file to its immediate parent and appends /spec.md, so a change to specs/<IR>/suite/generated/test_*.py looks for specs/<IR>/suite/generated/spec.md instead of specs/<IR>/spec.md. This causes generated-test changes to be rejected as “missing spec” before compilation and the advertised provenance/hash verification can run.
Agent Prompt
## Issue description
Changed files nested below a spec root are treated as though their immediate parent were the spec directory (by doing `dirname` + `/spec.md`), which fails for nested generated files and causes false missing-spec failures that prevent compilation and generated artifact provenance/hash verification.
## Issue Context
A valid generated output is located below `specs/<IR>/suite/generated/`, while its source of truth is `specs/<IR>/spec.md`. For each changed path, the workflow should locate the nearest owning ancestor containing `spec.md` under the `specs` directory (rather than blindly appending `spec.md` to `dirname`), and handle any grouping/metadata concerns separately from determining the owning spec directory.
## Fix Focus Areas
- .github/workflows/spec-dsl-gate.yml[83-110]
ⓘ Copy this prompt and use it to remediate the issue with your preferred AI generation tools
Comment on lines
+260
to
+262
| with open(out / "fanout-products.jsonl", "w", encoding="utf-8", newline="\n") as f: | ||
| for rec in fanout_records(report, card_id, spec_hash, base_sha): | ||
| f.write(json.dumps(rec, ensure_ascii=False) + "\n") |
There was a problem hiding this comment.
2. Append-only records are truncated 🐞 Bug ☼ Reliability
write_outputs() opens fanout-products.jsonl in "w" mode, deleting all existing fuel records whenever the output directory is reused. This contradicts the stated append-only contract and loses prior divergence and assumption records.
Agent Prompt
## Issue description
Every run truncates the append-only fanout product file and destroys records produced by earlier runs.
## Issue Context
Write new JSONL records using append-safe behavior consistent with the contract, while retaining deterministic overwrite behavior for the standalone report if desired.
## Fix Focus Areas
- pipeline/entropy/divergence.py[255-262]
- pipeline/INSTRUMENTS-B.md[46-48]
ⓘ Copy this prompt and use it to remediate the issue with your preferred AI generation tools
Comment on lines
+99
to
+106
| if raw.startswith("- "): # 新列表项（首字段随行） | ||
| cur = {} | ||
| acs.append(cur) | ||
| _fill(cur, raw[2:].strip(), source) | ||
| elif raw.startswith(" ") and cur is not None and ":" in raw: | ||
| _fill(cur, raw.strip(), source) | ||
| else: # 下一个零缩进键 → 节终止 | ||
| break |
There was a problem hiding this comment.
3. Indented ac lists rejected 🐞 Bug ≡ Correctness
extract_acceptance_criteria() treats a new acceptance-criteria item as valid only when - starts in column zero, but the repository’s canonical YAML spec format nests list entries under acceptanceCriteria: with two-space indentation. This causes acs to be parsed as empty and valid specs to be rejected, making the new compiler/PR gate fail on existing specs that the YAML-based validator accepts.
Agent Prompt
## Issue description
The DSL parser rejects valid `acceptanceCriteria` lists because `extract_acceptance_criteria()` only recognizes list items when `- ` begins at column zero, while repository specs (including fixtures/templates) use standard YAML nesting with two-space-indented list entries under `acceptanceCriteria:`.
## Issue Context
Repository specs place acceptance-criteria items beneath the YAML key and indent them (e.g., ` - id: ...`). The new parser currently branches only on unindented list markers, which results in `acs` being empty and triggers the parser’s empty-list failure, causing the new compiler/PR gate to reject specs that the canonical YAML-based spec validator (using `yaml.safe_load`) accepts.
## Fix Focus Areas
- pipeline/dsl/compile.py[83-116]
- pipeline/adversary/fixtures/intent-backstop/clean-spec.md[6-10]
- scripts/spec-check.py[94-126]
ⓘ Copy this prompt and use it to remediate the issue with your preferred AI generation tools
| as_union = sorted({it for m in as_maps for it in m.values()}) | ||
|
|
||
| convergence = round(100.0 * len(inter_keys) / len(union_keys), 2) if union_keys else 0.0 | ||
| mean_sim = round(sum(p["similarity"] for p in pairs) / len(pairs), 4) |
There was a problem hiding this comment.
4. Single skeleton crashes cli 🐞 Bug ☼ Reliability
A directory containing exactly one valid skeleton passes the non-empty directory validation but produces no pair combinations, causing the mean pair similarity (mean_sim/mean_pair_similarity) calculation to divide by zero. Because the CLI only catches EntropyError, this otherwise valid input triggers an uncaught ZeroDivisionError traceback instead of a controlled report or input error consistent with documented exit behavior.
Agent Prompt
## Issue description
A directory with exactly one valid skeleton currently passes input validation but generates zero pairwise combinations, leading the mean pair similarity calculation to divide by zero and crash with an uncaught `ZeroDivisionError`.
## Issue Context
The tool’s directory validation only rejects directories with zero `.md` files, but pairwise metrics inherently require at least two skeleton inputs. The CLI currently catches only `EntropyError`, so the division-by-zero escapes and violates the expected/documented controlled CLI exit behavior; the fix should either reject fewer than two skeletons via `EntropyError` or define an explicit, intentional behavior for the no-pairs case (and ensure it is covered by a CLI test).
## Fix Focus Areas
- pipeline/entropy/divergence.py[156-173]
- pipeline/entropy/divergence.py[213-214]
- pipeline/entropy/divergence.py[282-287]
ⓘ Copy this prompt and use it to remediate the issue with your preferred AI generation tools
Comment on lines
+159
to
+164
| body.append(f"def {fname}():") | ||
| body.append(f' """{ac["id"]}') | ||
| body.append("") | ||
| body.append(f" given: {_esc_doc(ac['given'])}") | ||
| body.append(f" when: {_esc_doc(ac['when'])}") | ||
| body.append(f" then: {_esc_doc(ac['then'])}") |
There was a problem hiding this comment.
5. Ac id breaks generated python 🐞 Bug ☼ Reliability
The raw AC ID is inserted into a triple-quoted docstring without escaping, although the other AC text is escaped. A valid YAML string such as id: 'AC-1"""' passes the DSL field checks but generates syntactically invalid Python; verify.py re-renders the same invalid text and reports it green.
Agent Prompt
## Issue description
Raw AC IDs are emitted inside Python triple-quoted docstrings and can terminate the literal, producing invalid generated test files.
## Issue Context
The existing spec validator only requires an ID to be a non-empty string, and the compiler already has an escaping helper for the other docstring fields.
## Fix Focus Areas
- pipeline/dsl/compile.py[119-125]
- pipeline/dsl/compile.py[145-166]
- scripts/spec-check.py[111-126]
ⓘ Copy this prompt and use it to remediate the issue with your preferred AI generation tools
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
变更
运行时证据
子代理自测 28 用例全绿（含对 IR-0004 真实 spec 的编译→verify 全链对拍、21 测试函数 pytest 实收、手改 hash 头不动也红的负向）；集成后本地复跑 py_compile+YAML 通过。
依据
IR-0004 rev6（PR .github#359）；ADR-0085 决策 7（门禁由 spec 自然携带、系统强制）。
Summary by CodeRabbit
新功能
文档
Cards: Cloudbird-Software/.github#329 Cloudbird-Software/.github#330