Skip to content
Agentmart
Submit a tool

lele-exec

Drive lele as an execution harness for an approved spec: feed the spec to lele, verify its output yourself, and iterate with session continuity (lele -p -s…

Skill by DixieIT

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/dixieit-lele-exec

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

Drive lele as an execution harness for an approved spec: feed the spec to lele, verify its output yourself, and iterate with session continuity (lele -p -s <id>) until the spec is met or 5 rounds fail. Trigger the moment a plan is approved — use this instead of implementing the plan directly.

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

Network hosts it talks to 1

  • codeload.github.com · found in install command

Files it touches

  • ~/.claude/skills/lele-exec

What's inside

Skill name
lele-exec
Description
Drive lele as an execution harness for an approved spec: feed the spec to lele, verify its output yourself, and iterate with session continuity (lele -p -s <id>) until the spec is met or 5 rounds fail. Trigger the moment a plan is approved — use this instead of implementing the plan directly.
License
MIT
Instructions
38 lines in SKILL.md · view on GitHub

Files (6)

  • SKILL.md
  • cli.md · reference/
  • goal.md · reference/
  • plugins.md · reference/
  • sessions.md · reference/
  • sub-agents.md · reference/

SKILL.md (first lines)

You have a spec. lele is your execution arm. You own acceptance — lele's "done" is a claim you verify.

## Loop structure

```
spec → lele -p "<spec>" → verify → if pass: done
                                    if fail: lele -p -s <id> "failure: ..." → verify → ...
```

- Max 5 rounds, then escalate to the user
- Each round: one `lele -p` call, then one verify step
- Use lele-goal (`create_goal`) to register the spec inside each lele session so the inner model stays anchored to it

## Step by step

1. **Prepare the call** — the spec is already the prompt. Prepend `create_goal({"objective": "<spec gist>"}); then implement` so lele-goal anchors the session. Append any shared contracts from scratch dir if sub-agents are needed.

2. **Call lele** — `lele -p -a <name> "..."` if a project defines a named agent (e.g. orchestrator), else plain `lele -p "..."`. Capture the session id from the last line.

3. **Verify** — run the app, run tests, check outputs against the spec. "Done" from lele is a claim, not a result. Use concrete commands with expected vs actual output.

4. **Iterate or land** — if verification passes: done, report. If fail: `lele -p -s <id> "<command run, expected vs actual, any errors>"` and loop.

5. **Escalate** — after 5 rounds of failure-to-verify, stop. Report what was attempted, what broke, and the best next hypothesis. Do not loop beyond 5 without a human call.

## lele-goal

`create_goal`/`update_goal`/`get_goal` tools, active by default (no config). Details: `reference/goal.md`.

## Sub-agent orchestration

Let lele's own `delegate`/`handoff` tools fan out work inside a single session. Only spawn separate `lele -p` calls when subtasks genuinely can't share one session (different working dirs, or the user explicitly wants split). If using a named agent (`-a`), check its `AGENTS.md` frontmatter first — named agents only get delegation tools if their frontmatter activates them.

## Reference docs

See `reference/` for full details on CLI flags, plugins, sessions, sub-agents, and lele-goal.

From DixieIT/skills, MIT.

Signals

Stars
0
Forks
0
Open issues
0
Last commit (this folder)
Jul 13, 2026 (3 months ago)
Last release
Jul 11, 2026 (3 months ago)

From GitHub (fetched Oct 24, 2026).

Details

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

More from DixieIT

Cite this page

lele-exec skill. Agent Mart. https://agentmart.dev/skills/dixieit-lele-exec (last updated 2026-10-24)