Skip to main content

Claude Code plugins for the agentic SDLC

58 skills and 17 agents in 13 plugins take work from workshop notes to a Jira backlog, and from a ticket to a reviewed change. Every non-trivial change starts from a plan you have approved.

Get it on GitHub ★ Read the docs
payments-api — claude
> Implement PROJ-123
⏺Skilltsh-product-engineering:orchestrating-feature-implementation
⎿ No plan for PROJ-123 yet, so one comes first
⏺Skilltsh-product-engineering:creating-implementation-plans
⎿ Wrote specifications/PROJ-123/payout-retry.plan.md · 2 phases, 4 tasks
⎿ Drafted payout-retry.verification.md · 3 scenarios you picked
The plan is ready. Read it, then choose:
❯ Execute the plan
  Wait — I want changes
⏺@tsh-product-engineering:software-engineer ×2 · Tasks 1.1 and 1.2, parallel group A
⎿ ✓ Definition of Done passed · plan checkboxes ticked
⎿ ✓ Phase 1 verification passed
⏺Final verification, in parallel
⎿ @tsh-product-engineering:code-reviewer ✓ approve · 0 blockers
⎿ @tsh-product-engineering:feature-verifier ✓ 3/3 scenarios pass
⏺PROJ-123 is ready for your review
⎿ What shipped, the review outcome and the verification results
>
specifications/PROJ-123/payout-retry.plan.md
1# PROJ-123 · Payout retry on bank timeout
2
3## Goal
4Retry a timed-out payout, at most 3 times
5
6## Technical Context
7- Test: npm test · Lint: npm run lint
8
9## Phase 1 — Payouts (parallel group A)
10- [x] Task 1.1 RetryPolicy with backoff
11 Files: retry.policy.ts (create)
12- [x] Task 1.2 Bank timeout metric
13 Files: timeout.metric.ts (create)
14Verification: npm test -- payouts
15
16## Phase 2 — Final verification
17- [x] Task 2.1 Code review
18- [x] Task 2.2 Functional verification
⎇ feature/PROJ-123specifications/PROJ-123/Claude Code
40%shorter lead time
300+engineers use it
17specialised AI agents
Freeopen source, MIT licence

Plugins for how each discipline works,
and for the context they all share

Product Management · 7 agentsSee the workflow →

Turns workshop material into a Jira-ready backlog

Workshoptranscript
Jiraexisting backlog
Figmadesigns and FigJam
PDFbriefs and specs
Confluencelinked pages
Codebasewhat exists today
JiraPROJ · Epic: PayoutsPushed only after you approve Gate 2
Retry payout on bank timeout
✓ 4 acceptance criteria
PROJ-201Story
Notify ops after 3rd failed attempt
✓ 3 acceptance criteria
PROJ-202Story
Show payout status to the merchant
✓ 5 acceptance criteria
PROJ-203Story
Audit log for every retry
✓ 3 acceptance criteria
PROJ-204Story
Cap retries to bank rate limits
✓ 2 acceptance criteria
PROJ-205Story
Admin view of failed payouts
✓ 4 acceptance criteria
PROJ-206Story
+ 18 more stories in 3 epics
Product Testing · 1 agentSee the plugin →

Writes end-to-end tests and proves them stable before handing them over

retry.spec.ts3/3 headless runs
1// PROJ-201 · criterion 1
2test('retries a payout',
3async ({ page }) => {
4 const po =
5 new PayoutsPage(page);
6 await po.open('PO-1042');
7 await po.retry();
8 await expect(po.status)
9 .toHaveText('Paid');
10});
11// ✓ user-visible locators
12// ✓ no waitForTimeout
13
Product Engineering · 6 agentsSee the workflow →

Plans the work, then hands it to specialised agents

ui-engineer✓ Figma gate passed
software-engineer✓ tasks done
code-reviewer✓ approve
feature-verifier✓ 3/3 scenarios
Main conversationorchestrating-feature-implementation skill
Platform Engineering · 2 agentsSee the plugin →

Builds and audits infrastructure, and never applies it unless you name the command

infrastructure-auditor · payments-api1 High
infra/iam.tf
12data "aws_iam_policy_document" "worker" {
13 statement {
14 actions = ["s3:*"]
15 resources = ["*"]
16 }
17}
18
19resource "aws_iam_role_policy" "worker" {
20}
High

Wildcard actions on every resource. Scope the policy to the bucket it reads.

Youdevops-engineer: plan validated, cost impact reportedterraform apply ← you run this
Context management · tsh-core and tsh-product-managementSee the workflow →

Keeps what the project knows where Claude reads it

CLAUDE.md
→
<project>-context
/tsh-core:initCLAUDE.md, rules, decision index
initializing-project-context✓ navigating-project-context cites every fact

58 skills of senior know‑how across our disciplines and stacks

Practice drawn from 160+ projects The Software House has delivered, packaged as Claude Code plugins. Claude loads the skills a task needs when the work matches their description, so you don't have to remember the commands.

Disciplines · 4 pluginsProduct Engineering · Product Testing · Product Management · Platform Engineering

33 skills

Stacks · 8 pluginsRuntime: Frontend, Node.js, Python, PHP, Serverless · Cloud: AWS, Azure, GCP

19 skills

Core · tsh-coreNeeded regardless of discipline or stack

6 skills

Wrote a skill your team relies on? Contribute it, and everyone who installs that plugin gets it with the next release, on their next /plugin update. To write one, start with authoring-claude-extensions.
Contribute a skill on GitHub →

Skills and agents to start with
+61 more
17 agents ship alongside them

Battle-tested on commercial projects

40%*shorter lead time on average
2–3×faster requirement gathering
Case study · US healthtech

“We delivered a full project with 1 developer in the time 3 would normally take.”

600tasks from the BA agent
30–40min per feature
Read the case study →

* Average across several AI agentic projects at The Software House, compared with the same teams before adoption.

Your discipline travels with you,
the stacks travel with the repo

Get it on GitHub ★
01

Add the marketplace

Run it inside any Claude Code session. It registers the catalog; nothing is installed yet.

Copy> /plugin marketplace add \
  TheSoftwareHouse/
  agentic-collections

▸ Core        1 plugin
▸ Disciplines 4 plugins
▸ Stacks      8 plugins
✓ Catalog registered
02

Install, then sign in

/plugin install tsh-core@tsh-agentic-collections, plus your discipline and the repo's stacks; run /reload-plugins if the install asks. Then /mcp signs you in to Jira and Figma, once per machine.

Claude Codev2.1 or newer
● Installed
JiraVia Atlassian MCP
● Connected
ConfluenceVia Atlassian MCP
● Connected
FigmaVia Figma MCP
● Connected
Playwright CLISeparate install
● On your machine
03

Run it on a ticket

Claude plans, builds and reviews. You approve the plan first and get the report at the end.

Implement PROJ-123
✓Plancreating-implementation-plans
✓Code and testssoftware-engineer
✓Review: approvecode-reviewer
✓Scenarios passfeature-verifier
Report: what shipped, review, verification

Want a team that already works this way?

Work with The Software HouseA self-managed team in 2 weeks: senior engineers who run these plugins on client projects.Talk to an engineer →Roll it out in your teamOur AI Enablement program covers training and support while your engineers adopt it.See AI Enablement →