AgentSkills

AI agent skills and tools for the full software testing lifecycle: requirement analysis, test planning, test design, execution, defect management, closure, and test automation.

This repository contains reusable agent skills, self-contained instruction sets that guide AI coding agents (Claude Code, GitHub Copilot, Cline, etc.) to perform specialized QA and testing tasks. What started as a single Test Plan Generator is now a full STLC skill pack

📦 Repository Structure

│ ├── stlc_manual_testing/ # STLC skill pack — 7 phases, 14 skills │ ├── 01_jira-requirement-analyzer/ # Phase 1 — Requirement Analysis │ ├── 02_testplan-creater-via-jira_LIVE/ # Phase 2 — Test Planning (flagship) │ │ ├── SKILL.md │ │ ├── context/How_to_Create_TestPlan_VWO.md │ │ ├── copilot/test-plan.prompt.md │ │ ├── references/requirement-checklist.md │ │ ├── scripts/fetch_jira.sh │ │ └── template/test-plan-template.md │ ├── 03-test-design/ │ │ ├── test-scenario-designer/ │ │ └── api-test-designer/ │ ├── 04-test-case-development/ │ │ ├── test-case-writer/ │ │ └── test-data-generator/ │ ├── 05-test-execution/ │ │ ├── automation-script-generator/ │ │ ├── regression-suite-selector/ │ │ └── test-execution-tracker/ │ ├── 06-defect-management/ │ │ ├── bug-reporter/ │ │ ├── bug-triage-assistant/ │ │ └── rca-analyzer/ │ └── 07-test-closure/ │ ├── test-coverage-analyzer/ │ └── test-closure-reporter/

---

## 🗺️ STLC Roadmap — which skill fires when

```mermaid
flowchart LR
    A["01 RequirementAnalysis"] --> B["02 TestPlanning"]
    B --> C["03 TestDesign"]
    C --> D["04 Test CaseDevelopment"]
    D --> E["05 TestExecution"]
    E --> F["06 DefectManagement"]
    F --> G["07 TestClosure"]

    A -.- A1[jira-requirement-analyzer]
    B -.- B1[test-plan-generator]
    C -.- C1[test-scenario-designerapi-test-designer]
    D -.- D1[test-case-writertest-data-generator]
    E -.- E1[automation-script-generatorregression-suite-selectortest-execution-tracker]
    F -.- F1[bug-reporterbug-triage-assistantrca-analyzer]
    G -.- G1[test-coverage-analyzertest-closure-reporter]

---

🧰 STLC Manual Testing Pack

Concept: One skill per STLC activity. Each skill is a SKILL.md with a routing description, a step-by-step workflow, an output contract, and guardrails, so any agent produces the same shape of artifact every time.

Why: QA work done ad-hoc by an agent drifts: fabricated acceptance criteria, unprioritized scenarios, bug reports missing repro steps. Encoding each activity as a skill makes the output repeatable and reviewable.

Q&A — why a skill pack instead of one mega-prompt?

• Q: When do I reach for a pack skill? A: When you hit that STLC phase. “Is this story ready?” fires the requirement analyzer; “file a bug for this” fires the bug reporter. Descriptions route automatically.
• Q: What does it replace? A: Copy-pasted prompt templates and tribal “how we write test plans here” knowledge scattered across docs.
• Q: What’s the gotcha? A: Skills stop at human review gates by design. The agent drafts; a human signs off. No skill in this pack marks anything final on its own.

Every skill follows the same frontmatter contract:

---
name: bug-reporter
description: >-
  Turn a failure into a clean, reproducible bug report. Use when a tester
  says "file a bug for this", "write up this defect", or describes
  something broken and wants it documented properly.
license: MIT
metadata:
  author: Rekha Govardhanan
  pack: stlc-manual-testing
---

flowchart TD
    S[Scan spec for brittle locators] --> C{Classify severity}
    C -->|HIGH| H["nth-child / .nth() / index"]
    C -->|MEDIUM| M[styling-class selectors]
    C -->|LOW| L[raw text matches]
    H & M & L --> D[Inspect live DOM]
    D --> R["Rewrite by priority:getByRole → getByLabel →getByText → getByTestId → scoped CSS"]
    R --> V{"Unique match?count() === 1"}
    V -->|Yes| P[Patch code + rewrite map]
    V -->|No| U[Flag UNVERIFIED for human]

---

🚀 Getting Started (flagship: Test Plan Generator)

Prerequisites

• Jira API access — a Jira account with API token access.
• jq (recommended) — pretty-printing JSON output from the fetch script.

  brew install jq   # macOS
  

1. Fetch a Jira Ticket

cd stlc_manual_testing/02_testplan-creater-via-jira_LIVE

export JIRA_BASE_URL="https://your-domain.atlassian.net"
export JIRA_EMAIL="[email protected]"
export JIRA_TOKEN="your-api-token"

./scripts/fetch_jira.sh VWO-49

Returns a structured JSON object with key, summary, type, priority, components, labels, description, links, and attachments.

2. Generate a Test Plan

Option A — Claude Code / Cline (via SKILL.md)

“Create a test plan for VWO-49 using the test-plan-generator skill”

The agent will:

1. Fetch the ticket via MCP or the fetch_jira.sh script
2. Analyze requirements against the gap-analysis checklist
3. Draft the test plan using the template
4. STOP for human review — the plan is never marked final without approval

Option B — GitHub Copilot (via prompt file)

---

🧪 Test Plan Generator — Workflow

┌─────────────────────────────────────────────────────┐
│                                                     │
│  1. FETCH the Jira ticket                            │
│     └─ Via Atlassian MCP / Jira REST API / Paste     │
│                                                     │
│  2. ANALYZE against the requirement checklist        │
│     └─ Flag gaps: missing ACs, ambiguous wording,    │
│        missing test data, env, a11y, security, etc.  │
│                                                     │
│  3. DRAFT the test plan                              │
│     └─ Scope & Objectives                            │
│     └─ Gaps & Questions for the author               │
│     └─ Test Scenarios (P0 / P1 / P2)                 │
│     └─ Test Data & Environment                       │
│     └─ Risks & Assumptions                           │
│     └─ Entry / Exit Criteria                         │
│                                                     │
│  4. STOP FOR HUMAN REVIEW  ←── MANDATORY GATE       │
│     └─ List assumptions made                         │
│     └─ List open questions                           │
│     └─ Ask: "Approve or edit before I continue"      │
│                                                     │
└─────────────────────────────────────────────────────┘

Guardrails

---

📋 Gap-Analysis Checklist

The references/requirement-checklist.md covers 5 dimensions:

---

📝 Test Plan Template

The template/test-plan-template.md produces plans with 6 sections:

1. Scope & Objectives — What’s in scope, out of scope, and the testing goal
2. Gaps & Questions for the Author — Table surfaced from gap analysis (⚠️ / ❌)
3. Test Scenarios — Prioritized (P0/P1/P2) with type, scenario, and AC mapping
4. Test Data & Environment — Data, flags, roles, permissions
5. Risks & Assumptions — What was assumed and what could go wrong
6. Entry / Exit Criteria — Gates for starting and completing testing

---

---

🔧 fetch_jira.sh — Script Reference

./scripts/fetch_jira.sh 

Environment variables:

Required tools: curl, jq

The script fetches the issue and returns a curated JSON object with: key, summary, type, priority, components, labels, fixVersions, description, links, attachments

---

---

🛡️ Security

• API tokens are read from environment variables only — never hardcoded
• The fetch_jira.sh script never stores or logs credentials
• All plans are DRAFT until a human approves — no automated sign-off

---

📄 License

MIT — See SKILL.md for details.

---

✍️ Author

Rekha Govardhanan