Skip to content
Agentmart
Submit a tool

implementation-planning

Plan one bounded implementation ticket before code: produce a reviewed technical design and executable consumer-behavior TDD plan for another agent.

Skill by jhonDoe15

IndexedMIT
Write a review

Install

Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.

npx agentmarket install skill/jhondoe15-implementation-planning

Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.

Get the weekly diff

The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.

We use your email only for this newsletter. Privacy

About

Plan one bounded implementation ticket before code: produce a reviewed technical design and executable consumer-behavior TDD plan for another agent.

Use cases: design, testing, agent-building

No reviews yet. Installed it? Reviews come from verified installers only. How reviews work

Permissions

What this skill can do and what it asks for, from its SKILL.md. Checked Oct 25, 2026.

Tools
No tools found
Secrets
None asked for
Install scripts
None found
Runs as
Node.js package

Network hosts it talks to 1

  • codeload.github.com · found in install command

Files it touches

  • ~/.claude/skills/implementation-planning

What's inside

Skill name
implementation-planning
Description
Plan one bounded implementation ticket before code: produce a reviewed technical design and executable consumer-behavior TDD plan for another agent.
License
MIT
Instructions
148 lines in SKILL.md · view on GitHub

Before installing

  • Ships executable scripts

Files (25)

  • SKILL.md
  • component.json · evals/
  • evals.json · evals/
  • cache-refresh.md · evals/fixtures/
  • export-timeout-contract.md · evals/fixtures/
  • fenced-lease-bounded.md · evals/fixtures/
  • pagination-consistency-contract.md · evals/fixtures/
  • payment-retry.md · evals/fixtures/
  • profile-column-migration.md · evals/fixtures/
  • simple-slug-normalization.md · evals/fixtures/
  • upload-cancellation-contract.md · evals/fixtures/
  • webhook-exactly-once.md · evals/fixtures/
  • outcome.json · evals/
  • package-closure.json · evals/
  • role.json · evals/
  • trigger.json · evals/
  • blueprint-review.md · references/
  • engineering-baseline.md · references/
  • engineering-compatibility.md · references/
  • engineering-data-migration.md · references/
  • engineering-reliability.md · references/
  • engineering-security.md · references/
  • evidence-and-authority.md · references/
  • implementation-blueprint.md · references/
  • contracts.test.js · test/

SKILL.md (first lines)

# Implementation Planning

Produce one **implementation blueprint** on the implementation ticket. The
blueprint's technical design and TDD test plan make implementation mechanical
for the lowest capable model while keeping design and risk decisions in
planning.

The blueprint is agent-primary and human-inspectable. Production artifacts stay
clear to both audiences.

## 1. Admit the planning work

Confirm the request targets one bounded implementation ticket and a pre-code
technical-design-plus-TDD outcome. Route decomposition, implementation,
debugging, one-off advice, broad product planning, and post-implementation
review to their owning flows.

Read [evidence and authority](references/evidence-and-authority.md) before
discovery. Apply its planning boundary and verify every required input or
record a named blocker with its owner and resolution condition.

*Complete when:* the ticket, requirements, repository, immutable base,
authority, prerequisites, constraints, exclusions, and publication
authorization are each verified or represented by a blocker.

## 2. Ground and discover

Preflight `engineering-guidance` and `architecture-review`; either missing
canonical Skill blocks dependent work. Invoke `architecture-review` against the
immutable current architecture and retain its brief as discovery evidence.
Preserve partial or failed results with review-recovery blockers; continue only
work supported by verified evidence. Read and disposition every item in
[the mandatory engineering baseline](references/engineering-baseline.md).
Classify work by residual
uncertainty and dispatch independent investigations only for the named gaps
defined in evidence and authority.

Reconcile returns centrally. Bind repository evidence to the immutable base,
close affected search paths, and record assumptions, contradictions, rabbit
holes, investigator provenance, and meaningful no-hit results.

*Complete when:* every requirement and affected path has authoritative,
base-bound evidence or a recovery blocker; every baseline item has a traceable
disposition or an applicability blocker; every
investigation has closed its scope or recorded a recovery blocker; and every
unresolved fact has an owner and gate.

## 3. Settle design authority

Apply the support boundary and design-authority gate from evidence and
authority. Select only behavior authorized by controlling sources and
capabilities verified by evidence. Keep implementation freedom bounded to the
smallest consistent design.

Represent each unresolved in-scope behavior or required capability with one
unranked decision envelope. Continue only work that is identical under every
option; place differing outcomes in a non-executable decision matrix.

*Complete when:* every implementation-shaping choice is settled with authority
and capability evidence or blocks its dependent design fields, scenarios,
commands, and slices without embedding a preferred option.

## 4. Draft the technical design

Use the retained `architecture-review` brief as the canonical architecture
scan, quality-claim, and persona evidence. Read [blueprint
review](references/blueprint-review.md) and [the blueprint
contract](references/implementation-blueprint.md) before drafting. Fill the
technical-design fields from settled evidence.

Reuse its ticket-bounded current-architecture scan; extend the six quality
claims to the proposed modules and blueprint as a whole. Draft current and
proposed behavior, ownership, interfaces, seams, lifecycle, effects, compatibility,
change map, durable rationale, alternatives, diagrams, and ordered
implementation slices. Link blocked fields to their decision envelopes.

*Complete when:* every applicable technical-design field is evidence-backed,
linked to a blocker, or has a justified `not-applicable`; every consumer
contract, invariant, lifecycle scenario, and changed module has an owner and
proof; and implementation receives no latent design work.

## 5. Review the design

Invoke `architecture-review` afresh against one numbered design revision, then
run blueprint review's Spec, Standards, specialist, and `needs-decision`
extensions against that same revision.

Gate unsettled design findings before comparing alternatives. Integrate
compatible corrections centrally, reread the whole artifact, and issue a new
revision only after verifying each correction landed consistently.

*Complete when:* every mandatory and triggered review on the same revision has
an evidence-bearing zero-finding `PASS`, complete findings, or an `INCOMPLETE`
record with recovery owner and condition; every exposed design choice is
settled or enclosed by a blocker. Any incomplete review keeps `needs-decision`.

## 6. Dispose findings and derive TDD

Apply blueprint review's lossless finding dispositions. Integrate every
credible in-scope correction into design, implementation slices, and proof;
preserve every source finding identity and closure.

Derive characterization locks and ordered vertical red-green slices from
settled consumer contracts, not private structure. Each slice names its
approved seam, setup, deterministic controls, stimulus, independent observable
assertion, smallest production responsibility, exact focused and regression
commands, expected red, green condition, and escalation stop.

Admit each proof command using the blueprint contract. Map every settled
requirement, interface promise, invariant, compatibility promise, and
applicable lifecycle, effect, or failure risk to a discriminating test,
justified non-test proof, evidenced `not-applicable`, or blocker.

*Complete when:* every source finding has one valid disposition and closure,
every settled proof obligation is traceable, every admitted slice has a full
red-green-regression loop, and no executable slice depends on a blocker.

## 7. Re-review and decide readiness

From jhonDoe15/skills, MIT.

More from jhonDoe15/skills

Signals

Stars
0
Forks
0
Open issues
16
Last commit (this folder)
Sep 6, 2026 (1 month ago)

From GitHub (fetched Oct 24, 2026).

Details

License
MIT
First seen
Oct 24, 2026
Last checked
Oct 24, 2026

More from jhonDoe15

Cite this page

implementation-planning skill. Agent Mart. https://agentmart.dev/skills/jhondoe15-implementation-planning (last updated 2026-10-24)