agent-task-compiler
Skill0 starsChanged 4 months agoMITView on GitHub
Compiles a vague project idea into agent-ready work - a sprint spec, a contract of testable assertions, and an ordered task list with acceptance criteria. Use when the user wants to hand a project to Codex, Claude Code, or coding agents, break down a build, or turn an idea into tickets/tasks.
In this repository5 other files it gives its agents.
- linkedin-signal-postsSkill
- skill-authorSkill
- skill-repair-loopSkill
- video-prompt-directorSkill
- voice-capsuleSkill
The file
We split this skill into the rules it gives an agent
1 rule. About 0 are common, 1 is distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 16
Existing code changes everything: If a repo already exists, task 1 is always "audit current state and reconcile this plan against it" — …
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Agent Task Compiler
Mission
Turn "build me X" into artifacts an agent can execute for hours without drifting: spec → contract → tasks. The contract is what gets graded, not the vibe of the original ask.
Workflow
- Preserve the boundary. Quote the user's original sentence verbatim at the top — it is the scope boundary everything else answers to.
- Write the spec: what done looks like (one paragraph, outcome not activity) · in scope · out of scope · constraints (stack, must-not-break, reuse-not-rewrite). If the idea contains several products, force it to one and park the rest — ask one batched question only if genuinely blocked.
- Negotiate the contract: 15–30 testable assertions. "Works well" is unfalsifiable; "searching 10k rows returns in under 2s" is a contract. Every assertion checkable by someone who didn't build it.
- Compile the tasks: ordered, each sized to a single agent session (≤ half a day). Per task: goal, files likely touched, acceptance criteria referencing contract IDs (A01, A02…), and what the next task depends on from it.
- Existing code changes everything. If a repo already exists, task 1 is always "audit current state and reconcile this plan against it" — never plan greenfield onto brownfield.
- Flag the human decisions. Any task involving payments, auth design, data deletion, or external communications gets a
NEEDS HUMANmarker with the specific decision required.
Boundaries
Decline to compile tasks for spam, mass-messaging, scraping that violates a platform's terms, or anything designed to deceive — offer the compliant version of the goal instead.
Output format
Original ask (verbatim)
Sprint spec
Contract
Numbered assertions A01–Axx.
Task list
Table: # · task · acceptance (contract IDs) · depends on · NEEDS HUMAN flags.
Parked
Anything cut from scope, in one line each, so nothing is silently lost.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

