Install
$ agentstack add skill-loomantix-claude-platform-task-packet ✓ 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
Task Packet Executor
You are given a Task Packet in markdown. Your job is to execute it end-to-end:
- Implement the described changes.
- Create/update the corresponding GitHub issue(s).
- Ensure tests (including E2E) pass and coverage is maintained.
- Open a PR wired to the issue.
- Report back with a clear summary and next steps for the user.
- Only consider the packet "done" after the user confirms it works and the issue is closed.
Phase 0: Parse & Validate the Task Packet
- Parse the argument as a markdown Task Packet following this structure:
# Task Packet: ...## Metadata## Context## Implementation Checklist## Testing & Coverage## GitHub Issue & PR## Observability & Post-Deploy Validation## Done When
- Extract key fields:
- Task name
- Area, type, risk, estimate, target branch
- Implementation checklist items
- Test commands and specific test files/specs
- GitHub issue instructions (existing vs create)
- PR requirements
- Acceptance criteria and “Done When” items
- Sanity-check:
- If critical info is missing (e.g., no indication of tests or no description of desired behavior), ask the user one concise clarification message before proceeding.
- If the packet is slightly under-specified but still executable, make pragmatic assumptions and call them out explicitly in your status summary.
Phase 1: Plan & Confirm
- Summarize in 3–6 bullet points:
- What you’re going to change
- Expected GitHub artifacts (issue + PR)
- Tests you’ll run (unit + E2E)
- Any obvious risks (migrations, infra, breaking changes)
- If Risk = high or the task involves:
- Database schema changes
- Infra/cluster changes
- Changes touching regulated-data handling (PHI, PII, PCI, secrets, etc.)
→ Ask for explicit confirmation from the user before proceeding.
- Once confirmed (or if clearly safe/low-risk), begin executing the packet.
Phase 2: Git & Environment Prep
Worktree Setup (New)
Before creating a branch or preparing the environment, create an isolated worktree for this Task Packet. This prevents conflicts between concurrent tasks.
- Determine Worktree Name
- Derive `` from the Task Packet name.
- Recommended format:
task-
Example: task-calendar-seed-fix
- Create or Reuse Worktree
- If the worktree directory does not exist:
``bash git worktree add worktrees/ ``
- If it already exists:
``bash git worktree unlock worktrees/ || true git -C worktrees/ pull --rebase ``
- Switch Execution Context
- All subsequent steps (branch creation, commits, tests, PR creation) must be performed inside this worktree:
``bash cd worktrees/ ``
- Create Task Branch
- Create the branch inside the new worktree, naming it as specified in the packet or derived from the task name (e.g.
feature/orbugfix/):
``bash git switch -c ``
- Environment
- Ensure dev environment is running, using commands from the packet (e.g., your repo's
just devrecipe,pnpm dev, ordocker compose up). - Run baseline tests specified in the packet to confirm the repo starts in a green state.
- Todo tracking
- Use TodoWrite (if available) to track each top-level checklist item as you complete it.
Phase 3: Implementation
- Follow the Implementation Checklist in order:
- Make high-confidence, minimal, idiomatic changes.
- Respect existing architecture, patterns, and conventions in this repo.
- Keep changes scoped tightly to the packet. Avoid “while I’m here” refactors unless they are:
- Trivial (e.g., typo fix) and
- Clearly beneficial and low-risk.
- Commits
- Prefer small, logical commits:
- Implementation
- Tests
- Minor fixes/cleanup
- Keep commit messages clear and imperative.
- Observability
- When instructed, add or update logs/metrics/traces:
- Use existing structured logging patterns and the canonical correlation fields your repo already uses (e.g., an
eventname plus per-domain ids). - Do not spam logs; prefer focused, useful events.
- If you discover inconsistencies between the Task Packet and the current code:
- Follow the packet’s intent but adapt to the real code.
- Note any deviations in the final status summary and in the PR description.
Phase 4: Testing & Coverage
Non-negotiable: you must run the tests specified in the Task Packet before calling it done.
- Unit / Integration Tests
- Implement or update tests in the paths specified.
- Ensure:
- Happy path is covered.
- Important edge cases are covered.
- At least one failure/exception scenario is covered if relevant.
- E2E Tests
- Implement or update the E2E specs from the packet (e.g., Playwright).
- Cover the core user flow(s) described.
- Run the E2E command(s) listed (e.g.,
pnpm exec playwright testor repo equivalent).
- Run All Required Commands
- Execute all test commands listed under “Commands to Run” in the packet.
- Fix any failures; rerun until green.
- Coverage
- If a coverage tool is configured, run it.
- Ensure:
- No meaningful drop in coverage in the files you touched.
- New branches/conditions in core logic are covered.
- If you can’t avoid a temporary drop, document why in the PR and mention it in your status summary.
Phase 5: GitHub Issue & PR
- Issue Handling
- If the packet specifies
existing:#: - Use that issue as the primary tracking item.
- If the packet specifies
create:new: - Create a new GitHub issue with:
- Title from “Proposed Issue Title”
- Labels from “Labels”
- Acceptance criteria from the “Issue Acceptance Criteria”
- You may use
gh issue createif available.
- Pull Request
- Open a PR from the feature branch to the target branch:
- Title: from the packet or derived from the task name.
- Description must include:
- Problem summary
- Implementation summary
- Testing summary (commands + key scenarios)
- Any deviations from the packet
Fixes #to auto-close the issue.
- Linking
- Ensure the GitHub issue and PR are properly linked.
- If auto-linking fails, add a comment on the issue linking to the PR.
Phase 6: Verification & Closure
- Local / Dev Validation
- Follow the “Manual Validation Steps” from the packet.
- Verify logs/metrics based on the “Logs / Metrics to Check” section.
- Status Report to User
- Post a concise but complete summary including:
- What changed (high-level)
- Files or areas heavily touched
- Tests run and their status
- Links/identifiers for:
- GitHub issue
- PR
- Any notable logs/dashboards (if applicable)
- Known limitations or follow-up items
- Wait for User Confirmation
- Do not manually close the issue yourself unless:
- It is clearly configured to close via
Fixes #, and - The user has confirmed it works, or this is a trivial/low-risk change.
- After Confirmation
- Ensure the GitHub issue is closed (via PR merge or manual action).
- Update any remaining TodoWrite items as complete.
- Suggest concrete follow-up issues if the packet uncovered new tech debt.
Response Style
When responding to the user (outside of terminal actions):
- Keep it structured and concise.
- Use checklists and bullet points.
- Focus on:
- What you did
- What you tested
- Where it lives (branch/PR/issue)
- What you recommend next
Do not restate the entire packet; only report on execution and deviations.
Source & license
This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.
- Author: loomantix
- Source: loomantix/claude-platform
- License: Apache-2.0
- Homepage: https://github.com/loomantix/claude-platform
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.