Install
$ agentstack add skill-samuelbostic29-claude-skills-plan-next ✓ 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
Plan Next: Execute the next phase of a saved plan
You are resuming work on a structured plan saved to a local file. The plan is ALREADY WRITTEN — do NOT create a new plan or enter plan mode. Your job is to find the next incomplete phase, implement it directly, update the plan file, and stop.
When to use this skill
- A saved plan file exists and you're resuming work on it — typically the first action of a fresh session.
- "Continue the plan" / "do the next phase" / "pick up where we left off".
When NOT to use this skill
- No saved plan exists — use
plan-saveto create one first; this skill executes, it doesn't author. - The user wants the plan changed — that's an edit to the plan file made with the user, not a phase execution. Never enter plan mode.
- The user wants a critique — use
plan-review. - "Run the whole plan" — the contract is one phase per invocation; if asked for more, say so and still do only the next phase (the user can re-invoke).
Steps
- Find the plan file. If an argument was provided, use that as the file path. Otherwise, search the current project directory for plan files (matching
*plan*.md,*PLAN*.md, or any markdown file containing## Implementation phases). If exactly one is found, confirm it with the user. If multiple are found, ask the user which one to use via AskUserQuestion — list them sorted by last-modified (most recent first) and default to the most recently modified file.
- Read the entire plan file.
- Understand the context. Read the Context section carefully. This is your only source of background information — there is no prior conversation history.
- Find the next open phase. In the "Implementation phases" section, find the first
###heading that is NOT prefixed withCOMPLETED. This is your current task.
- If all phases are COMPLETED, inform the user that the plan is fully implemented and stop.
- Implement the phase. Do the work described. Read existing code first before making changes. Use all available tools. If anything is unclear or blocked, ask the user rather than guessing.
- Update the plan file when the phase is done:
- Add an
#### Implementedsubsection at the end of the phase content with a brief, factual summary: files created/modified, key decisions made, any deviations from the plan. - Prefix the phase heading with
COMPLETED(e.g., change### 3. Create API endpointsto### COMPLETED 3. Create API endpoints).
- Stop. Do NOT proceed to the next phase. Report what was accomplished and stop.
Rules
- NEVER enter plan mode. Do NOT use EnterPlanMode or ExitPlanMode. The plan already exists — your job is to EXECUTE it, not re-plan it. Go straight to implementation.
- Only work on ONE phase per invocation.
- Execute phases in order — do not skip phases.
- If a phase is unclear or blocked, ask the user rather than guessing or improvising.
- The
#### Implementedsummary should be factual and concise — what files were touched, what changed, any deviations from the plan. - Do NOT modify the Context section or other phases.
- Do NOT move on to the next phase after completing one — mark it completed and stop.
- NO PLAN DEVIATIONS WITHOUT ASKING. If the planned approach seems difficult, infeasible, or suboptimal, you MUST stop and ask the user before changing the approach. Present the problem, explain why the plan seems hard, and propose alternatives — then let the user decide. NEVER unilaterally replace the planned approach with a different one, even if your alternative passes tests. The plan's approach was chosen for a reason (often architectural, not just functional), and silently deviating can undermine the entire purpose of the phase.
Source & license
This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.
- Author: SamuelBostic29
- Source: SamuelBostic29/claude-skills
- License: MIT
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.