feat(skills): add task compliance and traceability check skill

[svetkis]

---

name: Add Task Compliance Check Skill
about: Add feature-scope compliance skill for traceability mapping and scope creep detection
labels: enhancement, skill

Problem

Fast-ASDLC framework lacks a standardized mechanism to validate that a feature implementation matches its specification. Current gaps:

• No traceability: Acceptance criteria are not mapped to actual code changes
• Scope creep undetected: + lines unrelated to any AC slip into PRs unnoticed
• Regression blind spots: - lines that break existing functionality are not flagged
• No structured compliance report: Ad-hoc checks instead of systematic validation

Proposed Solution

Add a task-compliance-check skill under /.agents/skills/task-compliance-check/ that provides:

1. Traceability matrix: Map each AC to implementation files and test files
2. Scope creep detection: Flag + lines not traceable to any AC
3. Regression risk analysis: Flag - lines that break existing AC or delete tests
4. ANTI-HALLUCINATION protocol: Exact diff location, quoted + lines, spec paragraph reference
5. Routing decision: PROGRAMMER (missing AC) / QA (all good) / HUMAN_GATE (ambiguous)

Input

• git diff main...[feature-branch]
• .backlog/[feature].md
• docs/specs/[feature].md

Execution Phases

1. Parse AC from backlog/specs
2. Extract diff scope (changed files only)
3. Build traceability matrix
4. Detect scope creep and regression risk
5. Verify with evidence (quoted lines + spec refs)
6. Route to appropriate agent

Acceptance Criteria

• Skill file exists at /.agents/skills/task-compliance-check/task-compliance-check.md
• Skill defines input sources (diff, backlog, specs)
• Skill defines scope boundaries (only changed files)
• Skill includes traceability matrix format
• Skill includes scope creep detection rules
• Skill includes regression risk analysis rules
• Skill includes ANTI-HALLUCINATION verification steps
• Skill defines routing logic (PROGRAMMER/QA/HUMAN_GATE)
• Skill defines output format with traceability table
• Clean markdown, no lint errors
• LLM-friendly: concise, structured, minimal context window usage

Files Added

.agents/
└── skills/
    └── task-compliance-check/
        └── task-compliance-check.md

Methodology Alignment

• Spec-Driven: Validates that code follows specs before merge
• Everything-as-Code: Compliance procedure encoded as markdown
• Hexagonal Architecture: Checks ensure layer isolation is maintained
• LLM-Friendly: Structured phases reduce context window pressure

References

• Fast-ASDLC METHODOLOGY.md: Agentic compliance workflows
• Existing skills pattern: /.agents/skills/*/SKILL.md