Compare commits
5 Commits
895f551a0f
...
f781ba6566
| Author | SHA1 | Date |
|---|---|---|
|
|
f781ba6566 | |
|
|
0f708d0b89 | |
|
|
5fcdae02b5 | |
|
|
b8eeb78cff | |
|
|
b628eec9fd |
|
|
@ -1,7 +1,7 @@
|
||||||
<!-- if possible, run this in a separate subagent or process with read access to the project,
|
<!-- if possible, run this in a separate subagent or process with read access to the project,
|
||||||
but no context except the content to review -->
|
but no context except the content to review -->
|
||||||
|
|
||||||
<task id="_bmad/core/tasks/review-adversarial-general.xml" name="Adversarial Review (General)">
|
<task id="_bmad/core/tasks/review-adversarial-general.xml" name="Adversarial Review">
|
||||||
<objective>Cynically review content and produce findings</objective>
|
<objective>Cynically review content and produce findings</objective>
|
||||||
|
|
||||||
<inputs>
|
<inputs>
|
||||||
|
|
|
||||||
|
|
@ -43,15 +43,14 @@ Review `{git_discrepancies}` and create findings:
|
||||||
|
|
||||||
| Discrepancy Type | Severity |
|
| Discrepancy Type | Severity |
|
||||||
| --- | --- |
|
| --- | --- |
|
||||||
| Files changed but not in story File List | MEDIUM |
|
| Files changed but not in story File List | Medium |
|
||||||
| Story lists files but no git changes | HIGH |
|
| Story lists files but no git changes | High |
|
||||||
| Uncommitted changes not documented | MEDIUM |
|
| Uncommitted changes not documented | Medium |
|
||||||
|
|
||||||
For each discrepancy, add to `{context_aware_findings}`:
|
For each discrepancy, add to `{context_aware_findings}` (no IDs yet - assigned after merge):
|
||||||
|
|
||||||
```
|
```
|
||||||
{
|
{
|
||||||
id: "CAF-{n}",
|
|
||||||
source: "git-discrepancy",
|
source: "git-discrepancy",
|
||||||
severity: "...",
|
severity: "...",
|
||||||
description: "...",
|
description: "...",
|
||||||
|
|
@ -66,15 +65,14 @@ For EACH AC in `{acceptance_criteria}`:
|
||||||
1. Read the AC requirement
|
1. Read the AC requirement
|
||||||
2. Search implementation files in `{comprehensive_file_list}` for evidence
|
2. Search implementation files in `{comprehensive_file_list}` for evidence
|
||||||
3. Determine status: IMPLEMENTED, PARTIAL, or MISSING
|
3. Determine status: IMPLEMENTED, PARTIAL, or MISSING
|
||||||
4. If PARTIAL or MISSING → add HIGH severity finding
|
4. If PARTIAL or MISSING → add High severity finding
|
||||||
|
|
||||||
Add to `{context_aware_findings}`:
|
Add to `{context_aware_findings}`:
|
||||||
|
|
||||||
```
|
```
|
||||||
{
|
{
|
||||||
id: "CAF-{n}",
|
|
||||||
source: "ac-validation",
|
source: "ac-validation",
|
||||||
severity: "HIGH",
|
severity: "High",
|
||||||
description: "AC {id} not fully implemented: {details}",
|
description: "AC {id} not fully implemented: {details}",
|
||||||
evidence: "Expected: {ac}, Found: {what_was_found}"
|
evidence: "Expected: {ac}, Found: {what_was_found}"
|
||||||
}
|
}
|
||||||
|
|
@ -86,16 +84,15 @@ For EACH task marked [x] in `{tasks_with_status}`:
|
||||||
|
|
||||||
1. Read the task description
|
1. Read the task description
|
||||||
2. Search files for evidence it was actually done
|
2. Search files for evidence it was actually done
|
||||||
3. **CRITICAL**: If marked [x] but NOT DONE → CRITICAL finding
|
3. **Critical**: If marked [x] but NOT DONE → Critical finding
|
||||||
4. Record specific proof (file:line) if done
|
4. Record specific proof (file:line) if done
|
||||||
|
|
||||||
Add to `{context_aware_findings}` if false:
|
Add to `{context_aware_findings}` if false:
|
||||||
|
|
||||||
```
|
```
|
||||||
{
|
{
|
||||||
id: "CAF-{n}",
|
|
||||||
source: "task-audit",
|
source: "task-audit",
|
||||||
severity: "CRITICAL",
|
severity: "Critical",
|
||||||
description: "Task marked complete but not implemented: {task}",
|
description: "Task marked complete but not implemented: {task}",
|
||||||
evidence: "Searched: {files}, Found: no evidence of {expected}"
|
evidence: "Searched: {files}, Found: no evidence of {expected}"
|
||||||
}
|
}
|
||||||
|
|
@ -137,10 +134,10 @@ Present context-aware findings:
|
||||||
**Phase 1: Context-Aware Review Complete**
|
**Phase 1: Context-Aware Review Complete**
|
||||||
|
|
||||||
**Findings:** {count}
|
**Findings:** {count}
|
||||||
- CRITICAL: {count}
|
- Critical: {count}
|
||||||
- HIGH: {count}
|
- High: {count}
|
||||||
- MEDIUM: {count}
|
- Medium: {count}
|
||||||
- LOW: {count}
|
- Low: {count}
|
||||||
|
|
||||||
Proceeding to Phase 2: Adversarial Review...
|
Proceeding to Phase 2: Adversarial Review...
|
||||||
```
|
```
|
||||||
|
|
|
||||||
|
|
@ -75,74 +75,29 @@ If no baseline available, review current state of files in `{file_list}`:
|
||||||
|
|
||||||
### 2. Invoke Adversarial Review
|
### 2. Invoke Adversarial Review
|
||||||
|
|
||||||
<critical>Use information asymmetry: separate context from review</critical>
|
With `{diff_output}` constructed, invoke the review task. If possible, use information asymmetry: run this step, and only it, in a separate subagent or process with read access to the project, but no context except the `{diff_output}`.
|
||||||
|
|
||||||
**Execution Hierarchy (try in order):**
|
|
||||||
|
|
||||||
**Option A: Subagent (Preferred)**
|
|
||||||
|
|
||||||
If Task tool available with subagent capability:
|
|
||||||
|
|
||||||
```xml
|
```xml
|
||||||
<invoke-task subagent="true">
|
<invoke-task>Review {diff_output} using {project-root}/_bmad/core/tasks/review-adversarial-general.xml</invoke-task>
|
||||||
Review {diff_output} using {project-root}/_bmad/core/tasks/review-adversarial-general.xml
|
|
||||||
</invoke-task>
|
|
||||||
```
|
```
|
||||||
|
|
||||||
The subagent:
|
**Platform fallback:** If task invocation not available, load the task file and execute its instructions inline, passing `{diff_output}` as the content.
|
||||||
|
|
||||||
- Has FULL read access to the repository
|
The task should: review `{diff_output}` and return a list of findings.
|
||||||
- Receives ONLY `{diff_output}` as context
|
|
||||||
- Does NOT know story requirements, ACs, or intent
|
|
||||||
- Reviews code purely on technical merit
|
|
||||||
|
|
||||||
**Option B: CLI Fallback**
|
|
||||||
|
|
||||||
If subagent not available but CLI available:
|
|
||||||
|
|
||||||
```bash
|
|
||||||
# Pipe diff to adversarial review task
|
|
||||||
cat {diff_file} | claude --task {adversarial_review_task}
|
|
||||||
```
|
|
||||||
|
|
||||||
**Option C: Inline Execution**
|
|
||||||
|
|
||||||
If neither available, load `review-adversarial-general.xml` and execute inline:
|
|
||||||
|
|
||||||
1. Load task file
|
|
||||||
2. Adopt adversarial persona
|
|
||||||
3. Review `{diff_output}` with zero story context
|
|
||||||
4. Generate findings
|
|
||||||
|
|
||||||
### 3. Process Adversarial Findings
|
### 3. Process Adversarial Findings
|
||||||
|
|
||||||
Capture findings from adversarial review.
|
Capture findings from adversarial review.
|
||||||
|
|
||||||
**If zero findings returned:**
|
**If zero findings:** HALT - this is suspicious. Re-analyze or ask for guidance.
|
||||||
|
|
||||||
<critical>HALT - Zero findings is suspicious. Re-analyze or ask for guidance.</critical>
|
Evaluate severity (Critical, High, Medium, Low) and validity (Real, Noise, Undecided).
|
||||||
|
|
||||||
**For each finding:**
|
Add each finding to `{asymmetric_findings}` (no IDs yet - assigned after merge):
|
||||||
|
|
||||||
Assign severity:
|
|
||||||
|
|
||||||
- CRITICAL: Security vulnerabilities, data loss risks
|
|
||||||
- HIGH: Logic errors, missing error handling
|
|
||||||
- MEDIUM: Performance issues, code smells
|
|
||||||
- LOW: Style, documentation
|
|
||||||
|
|
||||||
Assign validity:
|
|
||||||
|
|
||||||
- REAL: Genuine issue to address
|
|
||||||
- NOISE: False positive (explain why)
|
|
||||||
- UNDECIDED: Needs human judgment
|
|
||||||
|
|
||||||
Create `{asymmetric_findings}` list:
|
|
||||||
|
|
||||||
```
|
```
|
||||||
{
|
{
|
||||||
id: "AAF-{n}",
|
source: "adversarial",
|
||||||
source: "adversarial-review",
|
|
||||||
severity: "...",
|
severity: "...",
|
||||||
validity: "...",
|
validity: "...",
|
||||||
description: "...",
|
description: "...",
|
||||||
|
|
@ -159,15 +114,15 @@ Present adversarial findings:
|
||||||
|
|
||||||
**Reviewer Context:** Pure diff review (no story knowledge)
|
**Reviewer Context:** Pure diff review (no story knowledge)
|
||||||
**Findings:** {count}
|
**Findings:** {count}
|
||||||
- CRITICAL: {count}
|
- Critical: {count}
|
||||||
- HIGH: {count}
|
- High: {count}
|
||||||
- MEDIUM: {count}
|
- Medium: {count}
|
||||||
- LOW: {count}
|
- Low: {count}
|
||||||
|
|
||||||
**Validity Assessment:**
|
**Validity Assessment:**
|
||||||
- Real issues: {count}
|
- Real: {count}
|
||||||
- Noise/false positives: {count}
|
- Noise: {count}
|
||||||
- Needs judgment: {count}
|
- Undecided: {count}
|
||||||
|
|
||||||
Proceeding to findings consolidation...
|
Proceeding to findings consolidation...
|
||||||
```
|
```
|
||||||
|
|
@ -186,7 +141,7 @@ Proceeding to findings consolidation...
|
||||||
|
|
||||||
- Diff constructed from correct source (uncommitted or commits)
|
- Diff constructed from correct source (uncommitted or commits)
|
||||||
- Story file excluded from diff
|
- Story file excluded from diff
|
||||||
- Subagent invoked with proper isolation (or fallback used)
|
- Task invoked with diff as input
|
||||||
- Adversarial review executed
|
- Adversarial review executed
|
||||||
- Findings captured with severity and validity
|
- Findings captured with severity and validity
|
||||||
- `{asymmetric_findings}` populated
|
- `{asymmetric_findings}` populated
|
||||||
|
|
@ -198,7 +153,7 @@ Proceeding to findings consolidation...
|
||||||
- Including story file in diff (breaks asymmetry)
|
- Including story file in diff (breaks asymmetry)
|
||||||
- Skipping adversarial review entirely
|
- Skipping adversarial review entirely
|
||||||
- Accepting zero findings without halt
|
- Accepting zero findings without halt
|
||||||
- Not using subagent when available
|
- Invoking task without providing diff input
|
||||||
- Missing severity/validity classification
|
- Missing severity/validity classification
|
||||||
- Not storing findings for consolidation
|
- Not storing findings for consolidation
|
||||||
- No explicit NEXT directive at step completion
|
- No explicit NEXT directive at step completion
|
||||||
|
|
|
||||||
|
|
@ -57,53 +57,39 @@ Keep the MORE DETAILED version:
|
||||||
- If adversarial finding has better technical detail → keep that
|
- If adversarial finding has better technical detail → keep that
|
||||||
- When in doubt, keep context-aware (has more context)
|
- When in doubt, keep context-aware (has more context)
|
||||||
|
|
||||||
Mark duplicates as merged:
|
Note which findings were merged (for transparency in the summary).
|
||||||
|
|
||||||
```
|
|
||||||
{
|
|
||||||
id: "CF-{n}",
|
|
||||||
merged_from: ["CAF-3", "AAF-2"],
|
|
||||||
kept_version: "CAF-3",
|
|
||||||
reason: "Context-aware version includes AC reference"
|
|
||||||
}
|
|
||||||
```
|
|
||||||
|
|
||||||
### 3. Normalize Severity
|
### 3. Normalize Severity
|
||||||
|
|
||||||
Apply consistent severity scale:
|
Apply consistent severity scale (Critical, High, Medium, Low).
|
||||||
|
|
||||||
| Severity | Icon | Criteria |
|
|
||||||
| --- | --- | --- |
|
|
||||||
| CRITICAL | RED | Security vuln, data loss, tasks marked done but not, broken core functionality |
|
|
||||||
| HIGH | ORANGE | Missing AC implementation, logic errors, missing critical error handling |
|
|
||||||
| MEDIUM | YELLOW | Performance issues, incomplete features, documentation gaps |
|
|
||||||
| LOW | GREEN | Code style, minor improvements, suggestions |
|
|
||||||
|
|
||||||
### 4. Filter Noise
|
### 4. Filter Noise
|
||||||
|
|
||||||
Review adversarial findings marked as NOISE:
|
Review adversarial findings marked as Noise:
|
||||||
|
|
||||||
- If clearly false positive (e.g., style preference, not actual issue) → exclude
|
- If clearly false positive (e.g., style preference, not actual issue) → exclude
|
||||||
- If questionable → keep with UNDECIDED validity
|
- If questionable → keep with Undecided validity
|
||||||
- If context reveals it's actually valid → upgrade to REAL
|
- If context reveals it's actually valid → upgrade to Real
|
||||||
|
|
||||||
**Do NOT filter:**
|
**Do NOT filter:**
|
||||||
|
|
||||||
- Any CRITICAL or HIGH severity
|
- Any Critical or High severity
|
||||||
- Any context-aware findings (they have story context)
|
- Any context-aware findings (they have story context)
|
||||||
|
|
||||||
### 5. Create Consolidated Table
|
### 5. Sort and Number Findings
|
||||||
|
|
||||||
|
Sort by severity (Critical → High → Medium → Low), then assign IDs: F1, F2, F3, etc.
|
||||||
|
|
||||||
Build `{consolidated_findings}`:
|
Build `{consolidated_findings}`:
|
||||||
|
|
||||||
```markdown
|
```markdown
|
||||||
| ID | Severity | Source | Description | Location |
|
| ID | Severity | Source | Description | Location |
|
||||||
|----|----------|--------|-------------|----------|
|
|----|----------|--------|-------------|----------|
|
||||||
| CF-1 | CRITICAL | task-audit | Task 3 marked [x] but not implemented | src/auth.ts |
|
| F1 | Critical | task-audit | Task 3 marked [x] but not implemented | src/auth.ts |
|
||||||
| CF-2 | HIGH | ac-validation | AC2 partially implemented | src/api/*.ts |
|
| F2 | High | ac-validation | AC2 partially implemented | src/api/*.ts |
|
||||||
| CF-3 | HIGH | adversarial | Missing error handling in API calls | src/api/client.ts:45 |
|
| F3 | High | adversarial | Missing error handling in API calls | src/api/client.ts:45 |
|
||||||
| CF-4 | MEDIUM | git-discrepancy | File changed but not in story | src/utils.ts |
|
| F4 | Medium | git-discrepancy | File changed but not in story | src/utils.ts |
|
||||||
| CF-5 | LOW | adversarial | Magic number should be constant | src/config.ts:12 |
|
| F5 | Low | adversarial | Magic number should be constant | src/config.ts:12 |
|
||||||
```
|
```
|
||||||
|
|
||||||
### 6. Present Consolidated Findings
|
### 6. Present Consolidated Findings
|
||||||
|
|
@ -115,10 +101,10 @@ Build `{consolidated_findings}`:
|
||||||
|
|
||||||
**Summary:**
|
**Summary:**
|
||||||
- Total findings: {count}
|
- Total findings: {count}
|
||||||
- CRITICAL: {count}
|
- Critical: {count}
|
||||||
- HIGH: {count}
|
- High: {count}
|
||||||
- MEDIUM: {count}
|
- Medium: {count}
|
||||||
- LOW: {count}
|
- Low: {count}
|
||||||
|
|
||||||
**Deduplication:** {merged_count} duplicate findings merged
|
**Deduplication:** {merged_count} duplicate findings merged
|
||||||
|
|
||||||
|
|
@ -126,16 +112,16 @@ Build `{consolidated_findings}`:
|
||||||
|
|
||||||
## Findings by Severity
|
## Findings by Severity
|
||||||
|
|
||||||
### CRITICAL (Must Fix)
|
### Critical (Must Fix)
|
||||||
{list critical findings with full details}
|
{list critical findings with full details}
|
||||||
|
|
||||||
### HIGH (Should Fix)
|
### High (Should Fix)
|
||||||
{list high findings with full details}
|
{list high findings with full details}
|
||||||
|
|
||||||
### MEDIUM (Consider Fixing)
|
### Medium (Consider Fixing)
|
||||||
{list medium findings}
|
{list medium findings}
|
||||||
|
|
||||||
### LOW (Nice to Fix)
|
### Low (Nice to Fix)
|
||||||
{list low findings}
|
{list low findings}
|
||||||
|
|
||||||
---
|
---
|
||||||
|
|
|
||||||
|
|
@ -79,7 +79,7 @@ The task should: review `{diff_output}` and return a list of findings.
|
||||||
|
|
||||||
Capture the findings from the task output.
|
Capture the findings from the task output.
|
||||||
**If zero findings:** HALT - this is suspicious. Re-analyze or request user guidance.
|
**If zero findings:** HALT - this is suspicious. Re-analyze or request user guidance.
|
||||||
Evaluate severity (Critical, High, Medium, Low) and validity (real, noise, undecided).
|
Evaluate severity (Critical, High, Medium, Low) and validity (Real, Noise, Undecided).
|
||||||
DO NOT exclude findings based on severity or validity unless explicitly asked to do so.
|
DO NOT exclude findings based on severity or validity unless explicitly asked to do so.
|
||||||
Order findings by severity.
|
Order findings by severity.
|
||||||
Number the ordered findings (F1, F2, F3, etc.).
|
Number the ordered findings (F1, F2, F3, etc.).
|
||||||
|
|
|
||||||
Loading…
Reference in New Issue