veripower
AGENTS.md10 starsChanged 53 days agoMITView on GitHub
VeriPower is a host-neutral IC design workflow plugin with independent stage skills. This file guides development of the plugin repository. Existing code and documentation are open to correction. - For repository changes, follow CONTRIBUTING.md for development and validation. - For scheduling, artifact provenance or acceptance changes, consult the architecture and the framework guide for the runtime model and implementation entry points. - For stage changes, inspect skills/<stage>/SKILL.md, its scripts and relevant references together. Shared machine interfaces live in framework/references/schemas/. -…
In this repository13 other files it gives its agents.
- .claude/CLAUDE.mdCLAUDE.md
- brainstormSkill
- design-flowSkill
- env-precheckSkill
- lint-cdcSkill
- power-analysisSkill
- rtl-designSkill
- simulation-planSkill
- simulationSkill
- simulation-triageSkill
- specificationSkill
- synthesisSkill
- timing-analysisSkill
The file
Files
VeriPower
VeriPower is a host-neutral IC design workflow plugin with independent stage skills. This file guides development of the plugin repository. Existing code and documentation are open to correction.
Task references
- For repository changes, follow CONTRIBUTING.md for development and validation.
- For scheduling, artifact provenance or acceptance changes, consult the architecture and the framework guide for the runtime model and implementation entry points.
- For stage changes, inspect
skills/<stage>/SKILL.md, its scripts and relevant references together. Shared machine interfaces live inframework/references/schemas/. - For installation and platform integration, use README.md and the relevant adapter guide.
Instructions needed to execute an installed skill belong in the skills and adapters, so execution works without this repository's AGENTS.md or the author's session context.
Runtime boundaries
framework/scripts/kernel.pyowns the module's append-onlyevents.jsonl, dispatch and outcome collection, and records decisions and signoff. Kernel commands, including queries, remove owner-write permission fromintent/to protect source materials from incidental writes.framework/scripts/rules.pydeclares stage inputs, directory roots and execution placement. These declarations define dependencies and legal repair owners, including the failed stage.skills/design-flow/SKILL.mddefines orchestration and applies the task's actual authorization to decisions. Platform adapters provide native tool integration and preserve host permissions.
signoff records acceptance when the task calls for it, after checking current proofs.
The record applies to the accepted evidence; readiness alone does not record acceptance.
Stage boundaries
Stages own their workdir artifacts and close through their own CLI; the kernel publishes
the outputs declared in each run's result.json. The module's intent/ tree contains the
original requirements and referenced authorities used to resolve discrepancies in derived documents.
Specification owns requirements, design choices and interface/clock boundaries. RTL design owns the module split and implementation. Verification derives expected behavior from independent requirements; implementation evidence may support diagnosis. Stage owners assess review findings and resolve demonstrated defects before acceptance.
brainstorm and env-precheck are pre-pipeline skills. They neither close stages nor write
kernel events.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
Reviews can't be read right now.
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.

