How do I install the Implement skill?

Run npx skillmds@latest add agentii-ai/implement in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.

Is Implement safe to use?

This skill has not completed SkillMD's automated safety review yet. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with Implement?

This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.

Is Implement free to use?

Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.

Who published Implement?

agentii-ai (@agentii-ai) published this skill. Their other Agent Skills are listed on their SkillMD profile.

Implement by agentii-ai — Agent Skill

Name: Implement
Author: agentii-ai

Implement

Execute research tasks — checklist soft gate, phase dispatch with explicit thesis_dir, budget enforcement (halt + approval card on overrun), skill_pin recording (version_hash content-hashed per skill directory), [x] written on completion.

agentii-ai b4720c1 2 files · 2.1 KB Updated Sep 17, 2026

File contents

agentii.implement

Before dispatch

Checklist soft gate (Q29): scan checklists/*.md; on unchecked items, stop and ask — the human may proceed (it is a soft gate, not a hard block).
Budget check (Q58): budget{max_tasks, max_retries_per_task} — exceeding halts with a Q39 approval card; never a silent overrun. The dispatcher checks budget first, max_in_flight second (Q82).
Resume matrix (Q56): absent → run; valid FR-090 → skip; corrupted → --resume; stale → leave for converge. The filesystem IS the checkpoint.
thesis_dir passed explicitly (Q37) — no singleton pointer; cwd fallback is human-only and journaled.

During dispatch

Cross-vertical handoff passes artifact paths, never content (Q13 rule 3).
skill_pin: {skill_name: version_hash} recorded before the first task (version_hash = content hash of SKILL.md + references/); a version change mid-thesis uses the new version and appends the new hash (Q57).
[x] written on completion — display-only; converge never reads it.

Invocation

python3 scripts/dispatch.py --thesis-dir theses/001-ai-semiconductors --task "NVDA recent-quarter default" --journal theses/001-ai-semiconductors/shards/run1.ndjson

Gate 4 (before mass dispatch, consequential — never delegable) shows budget and parallelism; it must attach a budget estimate.

Frequently asked questions