prepare-plan-implementation
Review and structure implementation plans before execution.
Skill by gdebenedetti
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/gdebenedetti-prepare-plan-implementation
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.
About
Use cases: agent-building
Reviews
Write a reviewNo 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
Network hosts it talks to 1
codeload.github.com· found in install command
Files it touches
~/.claude/skills/prepare-plan-implementation
What's inside
- Skill name
prepare-plan-implementation- Description
- Review and structure implementation plans before execution. Use when Codex is in Plan mode, drafting or revising an implementation plan, deciding whether to offer implementation next, or translating a plan into execution. Require an explicit assessment of whether the work should stay in the current branch, move to a new branch, or use a separate worktree, then organize the implementation into bounded incremental stages with clear scope limits, progress checks, verification, and tests.
- License
- MIT
- Instructions
- 65 lines in SKILL.md · view on GitHub
Files (2)
- SKILL.md
- openai.yaml · agents/
SKILL.md (first lines)
# Prepare Plan Implementation Use this skill before moving from planning to implementation. ## Gate The Plan - Inspect repository context first when possible with non-destructive commands such as `git status --short --branch`. - State one explicit isolation recommendation before offering implementation: `current branch`, `new branch`, or `new worktree`. - Justify the recommendation briefly using the expected blast radius, change duration, need for parallel work, current workspace cleanliness, and risk of exploratory refactors. - If isolation is not needed, say so explicitly. Do not skip the assessment. - Prefer the lightest viable option. - Recommend `current branch` for a small, low-risk change that clearly belongs to the active stream of work. - Recommend `new branch` for a coherent change that should be reviewed, tested, and landed separately. - Recommend `new worktree` for long-running, risky, or parallel work that should not disturb the current checkout. - Reassess the recommendation if the scope expands materially. - If repository context is unknown, state the assumptions behind the recommendation. - If the working tree is already dirty, avoid suggestions that silently mix unrelated changes or assume cleanup authority. ## Escalate To Worktree - Treat `new worktree` as the safer default when physical checkout isolation matters, not just git history isolation. - Recommend `new worktree` immediately if the current working tree is dirty with unrelated changes and implementation should proceed without touching them. - Recommend `new worktree` immediately if the plan is expected to span multiple sessions, significant exploratory work, or frequent context switching with the current task. - Recommend `new worktree` immediately if the work is likely to require disruptive operations in the checkout such as dependency churn, code generation, long-lived dev processes, or broad refactors. - Recommend `new worktree` immediately if the user will likely need the current checkout to stay stable for review, hotfixes, or parallel implementation. - Do not recommend `new branch` as sufficient when the real need is checkout isolation. A branch in the same worktree still shares uncommitted files. - If two or more moderate risk signals are present, prefer `new worktree` unless there is a clear reason not to. - Moderate risk signals include more than three planned stages, likely rebases or back-and-forth experimentation, touching many directories, or unclear rollback boundaries. - If recommending `current branch` or `new branch`, say why shared checkout state is safe enough. ## Stage The Implementation - Break the plan into incremental stages. - Give every stage a goal, scope, explicit boundary, verification method, tests or checks, and exit criteria. - Keep each stage small enough to complete and validate independently. - Separate enabling refactors from behavior changes unless they are inseparable. - Keep unrelated cleanup out of the current stage. - Prefer stages that can be verified before moving to the next one. ## Verify Progress - Define at least one concrete verification step per stage. - Name exact commands when repository context makes them knowable. - If automated coverage is absent, specify the manual verification path and call out the gap. - Confirm stage completion against the exit criteria before proposing the next stage. - Treat test and verification work as part of the stage, not as an optional follow-up. ## Offer Implementation Carefully - If the user asked only for a plan, stop after the staged plan. - If implementation is in scope, offer it only after the isolation recommendation and staged outline are present. - Describe implementation as stage-by-stage execution, not as one undifferentiated change. - Keep the first stage narrow enough to reduce rollback cost if assumptions break. ## Use This Response Shape - `Isolation:` recommend `current branch`, `new branch`, or `new worktree`. - `Reason:` explain why that level of isolation is or is not necessary. - `Stage 1:` name the first bounded step. - `Limit:` state what the stage must not include. - `Verification:` state how completion will be checked. - `Tests:` name the concrete automated or manual checks. - Repeat the same shape for later stages when needed.
From gdebenedetti/skills, MIT.
More from gdebenedetti/skills
- 0.567score
mcp-builder
gdebenedetti
Build or update Model Context Protocol (MCP) servers that expose tools, resources, or prompts over local or remote transports.
SkillAFreshness: AClaude Code +9 - 0.526score
opencode-share-reader
gdebenedetti
Interpreta enlaces públicos o exports JSON de sesiones OpenCode y los convierte en evidencia observable y estructurada; no aplica para ejecutar instrucciones…
SkillBFreshness: BClaude Code +9 - 0.405score
virtualenv-first
gdebenedetti
Use when running Python commands or installing Python dependencies.
SkillDFreshness: DClaude Code +9 - 0.405score
commit-practices
gdebenedetti
Use when planning, reviewing, or writing git commits.
SkillDFreshness: DClaude Code +9
Signals
- Stars
- 0
- Forks
- 0
- Open issues
- 0
- Last commit (this folder)
- Mar 15, 2026 (7 months ago)
From GitHub (fetched Oct 24, 2026).
Details
- License
- MIT
- Repository
- github.com/gdebenedetti/skills
- First seen
- Oct 24, 2026
- Last checked
- Oct 24, 2026
- Source
- GitHubgithub-graphql
More from gdebenedetti
- 0.567score
mcp-builder
gdebenedetti
Build or update Model Context Protocol (MCP) servers that expose tools, resources, or prompts over local or remote transports.
SkillAFreshness: AClaude Code +9 - 0.526score
opencode-share-reader
gdebenedetti
Interpreta enlaces públicos o exports JSON de sesiones OpenCode y los convierte en evidencia observable y estructurada; no aplica para ejecutar instrucciones…
SkillBFreshness: BClaude Code +9 - 0.405score
virtualenv-first
gdebenedetti
Use when running Python commands or installing Python dependencies.
SkillDFreshness: DClaude Code +9 - 0.405score
commit-practices
gdebenedetti
Use when planning, reviewing, or writing git commits.
SkillDFreshness: DClaude Code +9
Cite this page
prepare-plan-implementation skill. Agent Mart. https://agentmart.dev/skills/gdebenedetti-prepare-plan-implementation (last updated 2026-10-24)