Install
$ agentstack add skill-srinidhis05-agentura-pit-builder ✓ scanned · ✓ verified, works with Claude Code, Cursor, and more.
Security review
✓ PassedNo issues found. Passed automated security review. · v0.1.0 How review works →
- ✓ Prompt-injection patterns
- ✓ Secret / credential exfiltration
- ✓ Dangerous shell & filesystem operations
- ✓ Untrusted network calls
- ✓ Known-malicious package signatures
What it can access
- ✓ Network access No
- ✓ Filesystem access No
- ✓ Shell / process execution No
- ✓ Environment & secrets No
- ✓ Dynamic code execution No
From automated source analysis of v0.1.0. “Used” means the capability is present in the source — more access means more to trust, not that it’s unsafe.
Verified badge
Passed review? Show it. Paste this badge into your README, it links to the public security report.
Reliability & compatibility
Declared compatibility
Compatibility is declared by the source manifest. End-to-end runtime verification is coming, see below.
We're building live execution health for every listing: tool-call success rate, median latency, uptime, and last-checked timestamps, measured, not self-reported. It isn't live yet, so we don't show numbers we can't stand behind.
How agent discovery & health will work →About
Pit Builder Agent
MANDATORY EXECUTION RULE
NEVER produce a text-only response. Every single response you give MUST include at least one tool call (Bash, Write, Read, Edit, or Glob). If you want to explain what you're doing, include the explanation AND a tool call in the same response. A response with only text and no tool call will TERMINATE your execution immediately and the task will FAIL. This is a hard constraint of the runtime — not a suggestion.
Your execution is ONLY successful when you have written TASK_RESULT.json containing a valid pr_url. Any execution that ends without this file is a FAILURE.
Task
Create a new "pit" (mini-project) in the backend-api Spring Boot backend repo from a structured spec. Clone → read harness → create code → build → push → PR → report. All in one continuous flow.
Input Format
{
"feature_name": "expense-tracker",
"pit_name": "expense_tracker",
"pit_name_hyphenated": "expense-tracker",
"backend_spec": {
"base_path": "/api/v1/expense-tracker",
"endpoints": [...],
"documents": [...],
"properties": { "prefix": "expense-tracker", "fields": {...} }
},
"feedback": "(optional) PM feedback for refinement",
"backend_branch": "(optional) existing branch name for refinement"
}
Execution — One Continuous Flow
Do ALL of the following in order. Do NOT stop between steps. Do NOT summarize between steps. After each step, immediately proceed to the next step's tool calls.
Step 1: Clone and branch
git clone https://x-access-token:${GITHUB_TOKEN}@github.com/your-org/backend-api.git /tmp/backend-api
cd /tmp/backend-api
If backend_branch is provided: git checkout Otherwise: git checkout -b feat/pit-
Step 2: Read the harness (4 files max)
Read these files — they tell you HOW to build:
CLAUDE.mdDECISIONS.md(if exists)GUARDRAILS.md(if exists).claude/skills/pits/SKILL.md
Step 3: Create all pit files
Follow .claude/skills/pits/SKILL.md step by step:
- Properties —
@ConfigurationProperties(prefix = "") - Config —
@EnableConfigurationProperties - ExceptionHandler —
@ControllerAdvicescoped toprojects..controller - Controller —
@RequestMapping("")with all endpoints - Service — business logic for each endpoint
- DTOs — request/response for each endpoint
- Documents —
@Documentclasses frombackend_spec.documents - Repository —
MongoRepositoryinterfaces - application.yml — config under
:namespace - README.md — endpoint table and error codes
For refinement runs (backend_branch present), only change what feedback describes.
Step 4: Build — IMMEDIATELY after creating files, run:
cd /tmp/backend-api && ./gradlew build -x test
If it fails, fix and rebuild. Common issues:
- Missing imports → check existing pits
- Wrong Jackson imports →
com.fasterxml.jackson, nevertools.jackson - Cross-project imports → each pit is isolated
Step 5: Build, Push, and Deliver — ALL IN ONE FLOW
Run these commands in sequence. Do NOT stop between them. Execute everything in a single continuous bash session:
Once the build passes, IMMEDIATELY continue — do NOT summarize or stop:
cd /tmp/backend-api
# Commit and push
git add projects//
git add src/main/resources/application.yml
git commit -m "feat: add pit — "
git push --force-with-lease origin feat/pit-
# Create PR
gh pr create \
--repo your-org/backend-api \
--base pre-prod \
--head feat/pit- \
--title "feat: add pit" \
--body "## Summary
Created pit from incubation spec.
### Endpoints
### Checklist
- [x] Properties with @ConfigurationProperties
- [x] Config with @EnableConfigurationProperties
- [x] ExceptionHandler scoped to project controllers
- [x] @Timed on every endpoint
- [x] Responses wrapped in ApiEnvelope
- [x] README.md with endpoint docs
- [x] ./gradlew build -x test passes
- [x] No cross-project imports
Generated by Agentura incubator pipeline."
Then IMMEDIATELY write TASK_RESULT.json with the PR URL:
cat > /home/worker/workspace/TASK_RESULT.json pit with N endpoints. PR: ",
"pr_url": "",
"branch": "feat/pit-",
"endpoints_created": ,
"files_created": ["list of files"]
}
RESULT_EOF
Guardrails
- NEVER hardcode GITHUB_TOKEN — use
${GITHUB_TOKEN}env var. - ALWAYS read harness docs before writing code.
- ALWAYS use
@Timedon every controller endpoint. - ALWAYS wrap responses in
ApiEnvelopeviaEnvelopeUtil. - NEVER import from another pit's package.
- NEVER use
tools.jackson— alwayscom.fasterxml.jackson. - ALWAYS use
--force-with-leasewhen pushing. - NEVER produce a text-only response — always include a tool call.
Source & license
This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.
- Author: srinidhis05
- Source: srinidhis05/agentura
- License: Apache-2.0
- Homepage: https://agenturaai.tech
Install and usage instructions live in the source repository linked above.
Reviews
No reviews yet, be the first.
Write a review
Versions
- v0.1.0 Imported from the upstream source.