- Home
- Skills
- Prompt & Skill Engineering
- Brief to Approved Build Plan
More screenshots
Works with the AI Agents you already use
Brief to Approved Build Plan
Turn software briefs into approved specifications and implementation plans with mandatory human-in-the-loop checkpoints.
Free
Brief to Approved Build Plan
Example session with this skill installed
Plan a CSV export feature for our task management app.
Users should be able to export the currently filtered task list, including only the visible columns. If no tasks match, export a header-only CSV. Keep existing access permissions unchanged.
Create a specification with testable acceptance criteria. Stop for my approval before preparing the implementation plan. Do not write code.
- Read your context and instructions
- Compiled the brief to approved
Specification ready for review — spec-r1
Status: Awaiting your approval
Implementation plan: Not started
SCOPE
Export the currently filtered task list as CSV, using only the visible columns and preserving existing access permissions.
REQUIREMENTS
REQ-001 — Export only tasks matching the active filters.
REQ-002 — Include only visible columns, in their displayed order.
REQ-003 — Produce a header-only CSV when no tasks match.
REQ-004 — Preserve existing access permissions.
REQ-005 — Encode values containing commas, quotes or line breaks correctly.
ACCEPTANCE CRITERIA
AC-001 — Given a filtered list, when export is requested, the CSV contains all matching tasks and excludes non-matching tasks.
AC-002 — Given hidden columns, the CSV excludes those columns and preserves the displayed order of visible columns.
AC-003 — Given no matching tasks, the CSV contains column headers and no data rows.
AC-004 — A task the user cannot access is never included in the export.
AC-005 — Values containing commas, quotes or line breaks remain intact when the CSV is parsed.
ASSUMPTION TO CONFIRM
“Currently filtered task list” means all matching tasks the user can access, including those beyond the current page.
OUT OF SCOPE
Scheduled exports, additional export formats and changes to access permissions.
EVIDENCE LIMIT
No repository was supplied. Existing components and implementation details have not been inspected.
DECISION REQUESTED
Do you approve spec-r1, including the pagination assumption, or would you like changes?
Once approved, I will prepare an implementation plan mapping tasks and verification methods to these acceptance criteria. No code has been written.
Connects securely to your tools. The creator never sees your data.
About this skill
Turn a rough feature idea into a specification and build plan you have explicitly approved.
Brief to Approved Build Plan guides your AI agent through software planning with clear deliverables, human review checkpoints and revision tracking. It helps you agree on what should be built before implementation begins.
How it works
- Clarify the brief. The agent identifies the intended users, desired outcome, constraints and scope. It reuses information you have already provided and asks about decisions that materially affect the work.
- Create a testable specification. Your brief becomes a structured document with identifiable requirements, observable acceptance criteria, relevant failure cases, assumptions and explicit exclusions. When project files are available, the agent grounds its recommendations in the existing code and documentation.
- Wait for your approval. The agent presents the specification revision for review and stops before preparing the implementation plan. Requested changes produce a new revision; approval of an earlier version does not silently carry over.
- Build a traceable implementation plan. Once the specification is approved, the agent breaks the work into tasks with dependencies, affected components, requirement references and verification methods. Proposed file locations are distinguished from existing, inspected files.
- Review and hand off. You review the plan before it becomes the agreed baseline. The final handoff identifies the approved revisions, deferred decisions and first recommended implementation task. Coding remains a separate request. What makes this useful
- Human approval checkpoints: producing a document or passing a checklist does not count as your approval.
- Revision-aware tracking: changes to agreed requirements reopen the relevant decision and mark dependent plans as stale.
- Acceptance coverage: every acceptance criterion is mapped to implementation work and a concrete verification method.
- Clear evidence boundaries: inspected project facts, proposed designs, assumptions and unanswered questions remain distinguishable.
- Resumable planning: a project-specific journey document records current status and decision history without erasing earlier decisions. Included resources
- Specification template.
- Implementation-plan template.
- Journey and decision-history template.
- Review rules for specification quality, acceptance coverage and approval handling.
- A worked example showing approvals, requested changes and stale downstream plans. Typical use cases
Use it to plan a new software feature, clarify a client request, prepare a development handoff or scope a change in an existing repository. It is particularly useful when you want to review requirements before an agent starts coding.
Deliverables
With project file access, the agent maintains three Markdown documents:
- specification.md — requirements, acceptance criteria, constraints and open decisions.
- implementation-plan.md — tasks, dependencies, coverage and verification methods.
- journey.md — deliverable revisions, statuses, decision history and next action. In a chat-only environment, the documents are presented inline. Conversation and deliverables follow your chosen language.
Requirements and limitations
No additional package, API key or external account is required by the skill itself. Persistent deliverables require an agent with file-writing tools. Repository-aware planning requires access to the relevant project files.
This is an instructional workflow, not a technical enforcement system or an authenticated approval service. Approval records depend on available conversation evidence. The skill does not implement code, install tools or deploy software. Verification steps in a plan are proposed checks, not evidence that implementation tests have passed.
Example prompt
“Turn the following feature brief into a specification and implementation plan. Inspect the relevant project files, make assumptions explicit, and stop for my approval at each checkpoint. Here is my brief: …”
Planning only: does not implement code or deploy. Approval checkpoints rely on the agent following instructions, not technical enforcement. Saving deliverables requires file access; otherwise, output is provided in chat. New sessions need the previous documents and approval context to resume.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Free forever
No account required to browse
Trust & safety
Security scan
Security scan passed 10 days ago
- Free with an Agensi account
Requires
Permissions
File Scopes