agileflow 3.4.3 → 4.0.0-alpha.10

package/content/plugins/reviews/plugin.yaml

@@ -0,0 +1,14 @@
+id: reviews
+name: PR Reviews
+description: Structured pull request review covering security, logic, tests, and style.
+version: 1.0.0
+enabledByDefault: false
+cannotDisable: false
+depends: []
+provides:
+  skills:
+    - id: agileflow-pr-reviewer
+      dir: skills/agileflow-pr-reviewer
+  agents: []
+  hooks: []
+  templates: []

package/content/plugins/reviews/skills/agileflow-pr-reviewer/SKILL.md

@@ -0,0 +1,241 @@
+---
+name: agileflow-pr-reviewer
+version: 1.0.0
+category: agileflow/reviews
+description: |
+  Use when the user wants a structured review of a pull request, diff,
+  or code change. Covers security vulnerabilities, logic correctness,
+  test coverage gaps, API contract changes, breaking changes, and
+  code style. Produces a prioritised findings list and a final
+  recommendation: Approve, Request Changes, or Needs Discussion.
+triggers:
+  keywords:
+    - review this pr
+    - review my changes
+    - code review
+    - pr review
+    - check my code
+    - review this diff
+    - look at my code
+    - review before merge
+    - review this branch
+    - give me feedback on
+  priority: 50
+  exclude:
+    - review this document (not a code review)
+    - performance review (HR context)
+    - design review
+provides:
+  agents: []
+learns:
+  enabled: true
+  file: _learnings/pr-reviewer.yaml
+  maxEntries: 50
+depends:
+  skills: []
+  plugins: [core]
+---
+
+# AgileFlow PR Reviewer
+
+Systematic pull request review covering security, correctness, test coverage, API contracts, and style — with a prioritised findings list and a clear merge recommendation.
+
+## When this skill activates
+
+- User shares a diff, PR URL, branch name, or pastes code and asks for review
+- User says "review before I merge", "check this", or "give me feedback"
+- User wants a structured second opinion on their changes
+- User is implementing security-sensitive changes (auth, payment, file upload, SQL queries)
+
+## Opening discovery flow
+
+**When invoked without a diff, ask once to gather what's needed.**
+
+```xml
+<invoke name="AskUserQuestion">
+<parameter name="questions">[
+  {
+    "question": "What would you like me to review?",
+    "header": "Review source",
+    "multiSelect": false,
+    "options": [
+      {"label": "Paste the diff or changed files here (Recommended)", "description": "Copy from 'git diff' or paste the files you changed — I'll read them directly"},
+      {"label": "I'll give you the file paths — read them from disk", "description": "Tell me which files changed and I'll read them and the originals"},
+      {"label": "GitHub PR URL", "description": "Paste the PR URL and I'll fetch the diff via gh CLI"},
+      {"label": "Current staged changes", "description": "I'll run 'git diff --staged' to see what's about to be committed"}
+    ]
+  },
+  {
+    "question": "What kind of review do you need?",
+    "header": "Review focus",
+    "multiSelect": true,
+    "options": [
+      {"label": "Full review — security + logic + tests + style (Recommended)", "description": "Comprehensive check across all dimensions"},
+      {"label": "Security focus — OWASP top 10, injection, auth", "description": "Use for auth changes, API endpoints, file uploads, queries"},
+      {"label": "Logic and correctness", "description": "Race conditions, null handling, off-by-one errors, incorrect assumptions"},
+      {"label": "Test coverage", "description": "Are the right paths tested? Are assertions meaningful?"},
+      {"label": "API contract / breaking changes", "description": "Will this break existing consumers? Is the contract preserved?"}
+    ]
+  }
+]</parameter>
+</invoke>
+```
+
+## Review dimensions
+
+Run all dimensions for a full review, or the selected subset for a focused review.
+
+### 1. Security
+
+See `references/security-patterns.md` for the full checklist. Quick scan:
+
+- [ ] SQL injection / NoSQL injection (raw queries with user input)
+- [ ] Command injection (`exec`, `shell`, `subprocess` with user input)
+- [ ] Path traversal (`../../../etc/passwd`)
+- [ ] XSS (unsanitised user input rendered to HTML)
+- [ ] Insecure direct object reference (access control by user-controlled ID)
+- [ ] Authentication bypass (auth checks that can be skipped)
+- [ ] Sensitive data in logs, errors, or URLs
+- [ ] Hardcoded secrets or credentials
+- [ ] Cryptography: weak algorithms (MD5, SHA1 for passwords), improper key handling
+- [ ] CSRF on state-changing endpoints
+- [ ] Mass assignment (binding user input directly to DB models)
+
+### 2. Logic and correctness
+
+- [ ] Off-by-one errors in loops and index access
+- [ ] Null / undefined handling — does the code assume a value exists?
+- [ ] Race conditions — shared mutable state accessed concurrently
+- [ ] Error handling — are errors caught, logged, and propagated correctly?
+- [ ] Edge cases — empty collections, zero values, max values, unicode strings
+- [ ] Assumptions about data types — implicit type coercion, integer overflow
+- [ ] Conditional logic coverage — are all branches reachable and correct?
+- [ ] Async correctness — missing await, fire-and-forget, unhandled rejections
+
+### 3. Test coverage
+
+- [ ] Do the tests cover the new code paths?
+- [ ] Are there tests for the sad paths and edge cases?
+- [ ] Are assertions meaningful (not just `toBeTruthy`)?
+- [ ] Are mocks correct — do they mock at the boundary?
+- [ ] Would these tests catch a realistic regression?
+
+### 4. API contract and breaking changes
+
+- [ ] Does this change any public API endpoint signatures?
+- [ ] Are new required fields added to existing responses?
+- [ ] Are existing fields removed or renamed?
+- [ ] Is backward compatibility maintained for API consumers?
+- [ ] Are database migrations backward compatible (additive, not destructive)?
+- [ ] Are event schemas or message formats preserved?
+
+### 5. Code quality and style
+
+- [ ] Is the code readable and self-documenting?
+- [ ] Are variable and function names clear and consistent?
+- [ ] Is there duplication that should be extracted?
+- [ ] Are there comments explaining why (not what)?
+- [ ] Is the change within a reasonable size (< 400 LOC diff; flag if larger)?
+- [ ] Does it follow the project's established patterns?
+
+## Findings format
+
+Report each finding with a severity and actionable recommendation:
+
+```
+[P0] SECURITY — SQL Injection risk in user search
+  File: src/repositories/user-repo.js, line 47
+  Issue: User input interpolated directly into SQL query string
+  Fix: Use parameterised query: db.query('SELECT * FROM users WHERE email = $1', [email])
+  Reference: OWASP A03:2021 - Injection
+
+[P1] LOGIC — Missing null check before property access
+  File: src/services/order-service.js, line 112
+  Issue: order.customer.email accessed without checking if customer is null
+  Fix: Add guard: if (!order.customer) throw new OrderError('Customer not found')
+
+[P2] TESTS — No test for the refund failure path
+  File: tests/order.test.js
+  Issue: The error branch on line 89 of order-service.js has no test
+  Fix: Add test: 'throws RefundError when payment gateway returns 402'
+
+[P3] STYLE — Magic number should be a named constant
+  File: src/pricing/calculator.js, line 23
+  Issue: 0.15 used directly — unclear if this is a tax rate or discount
+  Fix: const DEFAULT_TAX_RATE = 0.15;
+```
+
+## Severity levels
+
+| Level    | Meaning                                                                    | Action required before merge                                       |
+| -------- | -------------------------------------------------------------------------- | ------------------------------------------------------------------ |
+| **P0**   | Security vulnerability or data corruption risk                             | Block merge — must fix                                             |
+| **P1**   | Logic error that will cause incorrect behaviour or crashes                 | Strong recommendation to fix before merge                          |
+| **P2**   | Missing test coverage for important path                                   | Fix before merge on critical features; acceptable on minor changes |
+| **P3**   | Code quality / style — won't cause a bug but makes code harder to maintain | Fix at author's discretion                                         |
+| **NOTE** | Observation, question, or suggestion with no required action               | Informational only                                                 |
+
+## Final recommendation
+
+End every review with one of:
+
+- **APPROVE** — No P0/P1 issues; P2/P3 are optional
+- **REQUEST CHANGES** — One or more P0 or P1 issues must be addressed
+- **NEEDS DISCUSSION** — Design concern that requires a conversation before code changes make sense
+
+```
+─────────────────────────────────────
+REVIEW SUMMARY
+─────────────────────────────────────
+Files reviewed: 8 | Lines changed: +247 / -89
+Findings: 1 P0, 2 P1, 3 P2, 1 P3
+
+VERDICT: REQUEST CHANGES
+
+Must fix before merge:
+  1. [P0] SQL injection in user-repo.js:47
+  2. [P1] Null dereference in order-service.js:112
+  3. [P1] Auth check missing on DELETE /api/users/:id
+
+Nice to have:
+  4. [P2] No test for refund failure path
+  5. [P3] Magic numbers should be constants
+
+After addressing P0 and P1, this is ready to merge.
+─────────────────────────────────────
+```
+
+## Self-improving learnings
+
+`_learnings/pr-reviewer.yaml` records:
+
+- Team's severity thresholds (e.g. P2 test coverage is a blocker on this team)
+- Recurring patterns to watch for in this codebase
+- Preferred fix patterns when issues are found
+- Merge criteria the team has established
+
+## Integration
+
+- **agileflow-audit** — use audit for pre-release sweeps or milestone quality gates; use pr-reviewer for individual PR merge gates; they're complementary not duplicates
+- **agileflow-debug** — when the reviewer finds a bug, hand off to debug for root cause analysis rather than patching inline
+- **agileflow-refactor** — when the review surfaces structural problems (long files, poor naming, deep coupling), route a follow-up story to refactor
+- **agileflow-test-writer** — when review finds missing test coverage, spawn test-writer to fill the gaps before approving the merge
+- **agileflow-accessibility** — when review touches UI components or interactive elements, accessibility check is part of the review criteria
+- **agileflow-performance** — when review touches query paths, rendering logic, or data processing, flag for performance review alongside the code review
+- **agileflow-docs** — if the PR changes public APIs or exports, verify docs were updated; pr-reviewer flags the gap, docs fills it
+- **agileflow-delivery** — pr-reviewer is a delivery gate; delivery orchestrates the overall release, pr-reviewer is one of its quality checkpoints
+- **agileflow-security** — invoke agileflow-audit with security dimension for auth changes, permission logic, or data handling code that warrants deeper security review
+
+## References
+
+| File                              | When to load                                                                 |
+| --------------------------------- | ---------------------------------------------------------------------------- |
+| `references/review-checklist.md`  | Full review — comprehensive checklist across all dimensions                  |
+| `references/security-patterns.md` | Security-focused review — OWASP patterns, injection, auth issues to scan for |
+
+## Workflows
+
+| File                           | When to follow                                                          |
+| ------------------------------ | ----------------------------------------------------------------------- |
+| `workflows/review-pr.md`       | Standard PR review — all dimensions, produces findings + recommendation |
+| `workflows/security-review.md` | Security-focused review for auth changes, API endpoints, data handling  |

package/content/plugins/reviews/skills/agileflow-pr-reviewer/references/review-checklist.md

@@ -0,0 +1,200 @@
+# PR Review Checklist
+
+**Load this when:** running a full review across all dimensions. Work through each section that is applicable to the change.
+
+---
+
+## Before you start: scope the review
+
+Read the PR description (or ask for it). Understand:
+
+1. **What is this change trying to do?** (feature, bugfix, refactor, chore)
+2. **What is the blast radius?** (one file, one service, cross-service, database schema)
+3. **What is the risk level?** (new endpoint, auth change, data migration, trivial copy change)
+
+Higher blast radius and risk = more thorough review needed.
+
+---
+
+## Section 1: Security
+
+Work through these for any change that handles user input, auth, database queries, or file operations.
+
+### Input handling
+
+- [ ] All user input is validated before use (type, length, format, range)
+- [ ] No raw SQL constructed by concatenating user input
+- [ ] No shell commands built from user input (`exec`, `spawn`, `subprocess`)
+- [ ] File paths from user input are sanitised (no `..` traversal)
+- [ ] HTML output is escaped (no raw `innerHTML = userInput`)
+- [ ] JSON parsed safely (no `eval()` on untrusted input)
+
+### Authentication and authorisation
+
+- [ ] Every state-changing endpoint has an auth check
+- [ ] Auth checks happen before expensive operations (not just before the response)
+- [ ] User can only access their own resources (no IDOR — Insecure Direct Object Reference)
+- [ ] Role checks are server-side (not just hidden in the UI)
+- [ ] Session tokens are invalidated on logout
+- [ ] Password reset tokens are single-use and time-limited
+
+### Data handling
+
+- [ ] No sensitive data (passwords, tokens, PII) logged to application logs
+- [ ] No sensitive data returned in error messages
+- [ ] No hardcoded credentials, API keys, or secrets in source code
+- [ ] Passwords hashed with bcrypt/argon2 (not MD5, SHA1, or SHA256 alone)
+- [ ] Tokens generated with cryptographically secure randomness
+
+### Transport and headers
+
+- [ ] HTTPS enforced (no plain HTTP fallback for sensitive operations)
+- [ ] CSRF protection on state-changing form submissions
+- [ ] Sensitive cookies have `HttpOnly`, `Secure`, and `SameSite` attributes
+
+---
+
+## Section 2: Logic and correctness
+
+### Null and undefined handling
+
+- [ ] No property access on a value that could be null/undefined
+- [ ] All function return values checked when they can return null/undefined/error
+- [ ] Optional chaining (`?.`) or explicit null checks before deep access
+- [ ] Array access guarded when index might be out of bounds
+
+### Async and concurrency
+
+- [ ] All Promises are awaited or properly handled with `.catch()`
+- [ ] No fire-and-forget async calls that silently fail
+- [ ] No shared mutable state modified concurrently without locks
+- [ ] No race condition between a read and a subsequent write
+- [ ] Database transactions used when multiple writes must be atomic
+
+### Error handling
+
+- [ ] All error paths handled explicitly (not just the happy path)
+- [ ] Errors are logged with enough context to diagnose (not just `console.error(err)`)
+- [ ] User-facing errors are generic (not leaking stack traces or internal details)
+- [ ] Errors propagate to the caller or are handled — not silently swallowed
+- [ ] Retry logic present for transient failures (network, DB connection)
+
+### Edge cases
+
+- [ ] Behaviour with empty input (empty string, empty array, zero) is correct
+- [ ] Behaviour at boundaries is correct (first/last item, min/max value, midnight)
+- [ ] Unicode strings handled correctly (emojis, RTL text, special characters)
+- [ ] Large inputs don't cause OOM or performance collapse
+- [ ] Pagination / cursors work correctly on empty result sets
+
+### Business logic
+
+- [ ] The change matches the acceptance criteria / ticket requirements
+- [ ] No "off-by-one" errors in loops, date calculations, or index slicing
+- [ ] No implicit assumptions about data ordering (unless ordering is guaranteed)
+- [ ] State transitions are valid (e.g. can't move from COMPLETED back to PENDING)
+
+---
+
+## Section 3: Test coverage
+
+### Coverage
+
+- [ ] New code paths have corresponding tests
+- [ ] Tests cover both the happy path and at least one sad path
+- [ ] Edge cases (null, empty, boundary values) have tests
+- [ ] Security-sensitive paths have explicit tests for unauthorised access
+
+### Test quality
+
+- [ ] Each test has at least one meaningful assertion (not just `toBeDefined()`)
+- [ ] Assertions are specific: `toBe(200)` not `toBeTruthy()`
+- [ ] Mocks are placed at the correct boundary (external deps only)
+- [ ] Tests don't depend on each other's state
+- [ ] Tests use descriptive names: `'returns 404 when user not found'` not `'test2'`
+
+### Test fragility
+
+- [ ] No fixed dates or time-sensitive assertions without fake timers
+- [ ] No hardcoded IDs or database sequences
+- [ ] No tests that rely on test execution order
+- [ ] No tests with `setTimeout` / `sleep` — use fake timers
+
+---
+
+## Section 4: API contracts and breaking changes
+
+### HTTP API
+
+- [ ] No required fields removed from request or response schemas
+- [ ] No fields renamed without a deprecation path
+- [ ] No HTTP status codes changed (e.g. 200 → 204 can break consumers)
+- [ ] New required request fields have defaults or a migration path for callers
+- [ ] Error response format consistent with existing endpoints
+
+### Database schema
+
+- [ ] Migrations are additive (add column, add table) — not destructive
+- [ ] `NOT NULL` columns have defaults or are added to empty tables
+- [ ] No column renames without code + data migration
+- [ ] Indexes added for new query patterns
+- [ ] Down migration exists (or impossibility noted)
+
+### Events / messages
+
+- [ ] Event/message schema changes are backward compatible
+- [ ] Consumers won't break if a new field is added
+- [ ] No event type renames without a transition period
+
+---
+
+## Section 5: Performance
+
+- [ ] No N+1 queries (loop that issues a DB query per iteration)
+- [ ] Large result sets are paginated — no `SELECT *` without `LIMIT`
+- [ ] Expensive operations aren't synchronous on the request thread (use queues)
+- [ ] Caches are invalidated correctly when underlying data changes
+- [ ] No unnecessary database round-trips (can be combined into one query)
+- [ ] Indexes exist for new filtering/sorting patterns
+
+---
+
+## Section 6: Code quality
+
+### Readability
+
+- [ ] Variable and function names clearly express intent
+- [ ] Functions do one thing (single responsibility)
+- [ ] Complex logic has a comment explaining _why_ (not _what_)
+- [ ] No dead code (commented-out blocks, unused variables, unreachable branches)
+
+### Duplication
+
+- [ ] No copy-paste of code that already exists elsewhere
+- [ ] New utility functions placed in a shared location if used in >1 place
+- [ ] Constants defined once, not repeated as magic numbers/strings
+
+### Change size
+
+- [ ] PR is reviewable (< 400 LOC diff as a guideline)
+- [ ] If larger: is it a mechanical change (rename, format)? Is it justified?
+- [ ] Each PR does one thing (not a feature + refactor + chore bundled together)
+
+---
+
+## Section 7: Documentation
+
+- [ ] New public functions/methods have JSDoc / docstrings
+- [ ] README updated if setup instructions changed
+- [ ] API documentation updated if endpoint signatures changed
+- [ ] Migration guide added if there are breaking changes
+- [ ] `CHANGELOG.md` entry added if this is a user-facing change
+
+---
+
+## Checklist usage
+
+For a standard review: work through Sections 1–4 for every PR.
+For a security review: focus on Section 1, run `workflows/security-review.md`.
+For a performance review: focus on Section 5.
+For a refactor-only PR: focus on Sections 2, 3, 4, 6.