Skills DirectorySkills Directory
SkillsLearnSecurityCategoriesDocsBlogPro
Sign InSubmit Skill
Skills Directory

Security-tested agent skills for Claude, coding agents, and AI workflows.

Directory

  • Browse Skills
  • All Skills A–Z
  • Claude Skills
  • Claude Code Skills
  • Agent Skills
  • Categories
  • Authors
  • Submit a Skill

Learn

  • Learn Hub
  • Install Claude Skills
  • Write SKILL.md
  • Skills vs MCP
  • Directories Compared

Security

  • Security
  • Methodology
  • Secure Claude Skills
  • Security Badges
  • Chrome Extension
  • Skill Manager

Company

  • About
  • Community
  • Blog
  • API Docs
  • Advertise

2026 Skills Directory. All rights reserved.

ProTermsPrivacyRefunds
Back to skills

Implement

ASecurity

Execute a published Linear issue by work type, with status tracking, test-first verification, current documentation, and a verified handoff.

3 stars
0 votes
0 copies
1 views
Added 9/23/2026
ai-agentsgorefactoringcode-reviewgitsecuritydocumentation

Security Analysis

A100/100

Pro scans all 6 files and shows the line behind each finding

Scanned 10/7/2026

$npx -y skills add Firzus/agent-skills --skill implement --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Implement?

Add the live security badge to your README — it updates automatically with every re-scan.

Security grade badge for Implement
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/firzus-implement/badge)](https://www.skillsdirectory.com/skills/firzus-implement)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
Files
SKILL.md
---
name: implement
description: Execute a published Linear issue by work type, with status tracking, test-first verification, current documentation, and a verified handoff.
disable-model-invocation: true
---

# Implement a verified change

Deliver one coherent, reviewable outcome. Tests, context, and affected documentation
belong to the change. This skill runs only on a published Linear issue; a request
without an issue goes to `triage` first: say so and stop. Invoking this skill on a
Linear issue authorizes, for that issue: status changes, commits, pushing a prototype
branch, publishing a Research issue's report as a Linear document attached to it
and closing the issue when its research completes, and closing a Prototype or
Interview issue after the user validates its result. For a Task or
Bug, ask the user to validate opening the PR once the work is verified; that
validation authorizes the push and a non-draft PR. It does not authorize
unrelated work, merge, or deployment.

## 1. Read the work and route it

1. Read the request, project instructions, current repository state, and existing
   work record. For a Linear issue, include comments, linked decisions, acceptance
   criteria, and native blocking relations.
   For a Task or Bug, record the starting revision and scoped local changes before
   any skill handoff or edits, so later review can distinguish this task's changes
   from existing work.
2. Apply [intake and handoff](references/intake-and-handoff.md) in its order: check
   the status, accept or refuse the issue, move it to In Progress, route it by its
   Type label, and verify prerequisites, prototype version, and context changes. A
   Prototype, Research, or Interview issue follows its route there instead of the
   rest of this workflow; a Bug goes through `debug` and resumes at step 4.
3. Trace the affected behavior through callers, contracts, configuration, and tests.
   Separate unrelated local work and pre-existing failures.
4. Resolve discoverable facts locally. For consequential open choices or conflicts,
   resolve them with `interview` when they stay inside the issue's scope, otherwise
   retain an out-of-scope finding in the conversation. Publish it as a Need triage
   issue only with approval of its content and destination, or explicit existing
   authorization covering that creation. Pause only the dependent work.

**Done:** scope, prerequisites, authorization, and independent acceptance evidence
are clear. Otherwise report the precise blocker and continue only independent,
authorized work. In read-only or planning mode, retain a plan rather than edit.

## 2. Prepare the change and verification

1. Preserve the prepared branch/worktree and unrelated changes. Identify existing
   components and dependencies to reuse; avoid speculative restructuring.
2. Establish the relevant baseline and map each acceptance criterion to a focused
   check, including consequential adverse cases for security, privacy, money,
   destructive operations, or public compatibility.
3. Choose the smallest coherent implementation sequence and the agreed delivery
   boundary. Read [delivery](references/delivery.md) before Git publication or
   Linear updates.

**Done:** baseline and checks identified, work isolated, delivery expectations explicit.

## 3. Implement with tests and documentation

For each production behavior:

1. Write a focused test at an existing caller-facing boundary.
2. Run it and inspect the failure: it must demonstrate the missing behavior,
   not a broken fixture, dependency, or command.
3. Make the smallest correct change using established components and contracts.
4. Run the test again; refactor while green, then move to the next behavior.

| Work shape | Verification |
| --- | --- |
| Change to code behavior or user-visible output, including displayed text | Test first when a test boundary exists for it |
| No test boundary exists, or test-first is otherwise unsuitable | State why and use relevant substitute evidence; avoid infrastructure solely for ceremony |
| Refactoring | Establish and preserve a passing behavioral baseline |
| Documentation only | Verify sources, claims, and links; no artificial code change or failing test |

- Deliver the accepted context delta using [intake and handoff](references/intake-and-handoff.md#deliver-context-with-the-change).
- When a change affects a system's overview or an explanation not adequately
  conveyed by code and comments, use
  [system documentation](references/system-documentation.md).
  Use that reference for documentation audits and retirement as well.
  Implementation changes alone do not require expanding a system page.
- Keep comments beside verified, non-obvious rationale or caller obligations.
  Explain workaround sources and removal conditions; update stale affected comments.
  Avoid narration, decorative labels, vague TODOs, and disabled code.
- A new consequential decision returns to clarification; ordinary internal choices
  remain with implementation. Research/prototype needs do not expand scope silently.

**Done:** each behavior has failing/passing evidence or a justified substitute;
code, context, contracts, and affected documentation agree.

## 4. Verify the complete result

1. Run [the review-correction loop](references/review-loop.md) using `code-review`.
   Reviewers inspect; implement verifies findings, corrects confirmed problems,
   and requests targeted follow-up. Keep optional cleanup out of the change.
2. Run focused checks and required project checks. Fix failures caused by the
   change; report pre-existing failures without broadening the task.
3. For user-visible behavior, exercise the ordinary integrated entry point,
   interactions, and relevant narrow or target-device layout. Keep the review
   surface available through supported tools.
4. Identify prototype, simulation, or live data. Distinguish inspected behavior,
   executed checks, and unavailable runtime evidence.
5. After a visual bug fix or when a recorded demo is requested, use the available
   `video-report` skill to capture and review evidence; its capture rules stay there.
   Video supplements tests, not replaces them. If unavailable, report the gap;
   it blocks delivery only when video is required acceptance evidence. Keep recordings
   local unless their publication is authorized.

**Done:** the review loop's stopping conditions hold, and acceptance evidence covers
the connected result and relevant failure cases.
Missing required verification remains a delivery limitation, not a claimed pass.

## 5. Deliver and report

1. Follow [delivery](references/delivery.md) for authorized commits, PRs, reviews,
   and meaningful Linear updates.
   For a Bug, the PR description and an issue comment carry the recap: symptom,
   cause and its evidence, fix, regression test and checks, video evidence when
   step 4 produced it, and remaining uncertainty.
2. Preserve the existing record; link artifacts and evidence rather than creating
   another backlog. Report acceptance, integration, and deployment separately.
3. Return the result, affected files, checks, limitations, and next owner/action.
   When video evidence was produced, include its supported preview or absolute local
   path, the demonstrated scenario and observed result, and any review limitations.
   A newly prepared issue or broader goal needs a separate request, not automatic
   continuation into the next feature.

**Done:** the agreed completion boundary is met and required updates are verified.
Otherwise report the completed local work and exact pending delivery operations.

Attribution

FirzusFirzus
View sourceSee grades on GitHubMore from Firzus →
SSkills DirectorySkills Directory

Ship a skill? Prove it's safe.

Free 120-pattern security scan, letter grade, and an embeddable README badge.

Submit a skill

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments (0)

No comments yet. Be the first to comment!

SSkills DirectorySkills Directory

Ship a skill? Prove it's safe.

Free 120-pattern security scan, letter grade, and an embeddable README badge.

Submit a skill

Related Skills

Caveman

Terse caveman voice: answer first, fluff gone, every technical fact kept. Use for /caveman, "caveman mode", "talk like caveman", "be brief", "less tokens". Stays on until "stop caveman" or "normal mode".

1100021 votes

Hyperplan

Adversarial multi-agent planning skill. Self-orchestrates 5 hostile category members (unspecified-low, unspecified-high, deep, ultrabrain, artistry) via team-mode for ruthless cross-critique debate, distills only the defensible insights, then MANDATORILY hands the distilled insight bundle to the `plan` agent for executable plan formalization. Use when planning needs maximum rigor and surfacing of weak assumptions, blind spots, and over-engineering. Triggers: 'hyperplan', 'hpp', '/hyperplan', ...

698461 votes

Writing Skills

Create and manage Claude Code skills in HASH repository following Anthropic best practices. Use when creating new skills, modifying skill-rules.json, understanding trigger patterns, working with hooks, debugging skill activation, or implementing progressive disclosure. Covers skill structure, YAML frontmatter, trigger types (keywords, intent patterns), UserPromptSubmit hook, and the 500-line rule. Includes validation and debugging with SKILL_DEBUG. Examples include rust-error-stack, cargo-dep...

3931 votes

Mcp Code Execution

Routes multi-tool workflows through MCP servers for large datasets and pipelines. Use when Bash tool overhead is limiting throughput on data-heavy tasks.

3421 votes

catchup

Recovers the conversation and failed tool calls of a previous Codex, Amp, Claude Code, Antigravity, Cline, Copilot CLI, Cursor, DeepSeek Harness, Grok Build, Kimi, OpenCode, Pi Agent, or ZCode session. Use when the user says "catch up", "what did the last session do", "get me up to speed", "I switched agents", asks to recover/summarize a previous session before continuing, or asks to diagnose or report a catchup failure. Do NOT use for the current conversation, git history, or any non-agent log.

741 votes
View all in ai-agents →