Run parallel internal and external spec reviews until convergence.
npx skills add https://github.com/JKHeadley/instar --skill spec-converge
Please help me install this Agent Skill. Skill: spec-converge Source: https://github.com/JKHeadley/instar/tree/main/skills/spec-converge Command: npx skills add https://github.com/JKHeadley/instar --skill spec-converge
This Skill prevents implementation from starting on a spec that still has hidden risks, unresolved decisions, or missing review coverage. It turns a draft spec into a converged spec by forcing multiple perspectives to examine it until the design is ready to build.
It runs structured reviews from security, scalability, adversarial, integration, decision-completeness, and lessons-aware perspectives, then folds in external cross-model feedback when available. It updates the spec to address findings, repeats the review cycle until convergence, and writes the review metadata and report back into the repository. Use it when you need a spec hardened before development, especially for changes that affect agent behavior, state, deployment, or user-facing workflows.
Run this skill on a spec file path to review it, resolve the findings, and generate a converged report before any implementation begins.
High-intent search queries and answers about installing and using this skill.
Spec review for security and integration risks requires running parallel internal and external checks until convergence. This skill applies security, scalability, adversarial, and integration validations to resolve hidden risks and open questions before implementation begins.
Spec convergence is the process of hardening a draft design by folding in cross-model feedback and lessons-aware validation. You need it when a spec affects agent behavior, state, deployment, or user-facing workflows and requires frontmatter tagging before implementation.
Yes, cross-model validation and scalability checks are supported. The convergence process runs structured reviews from multiple perspectives including security, adversarial, integration, and lessons-aware angles, folding in external cross-model feedback when available.
To ensure decision-completeness, run review cycles that identify unresolved decisions and open questions within the spec. The convergence process forces multiple perspectives to examine the document until all open questions are resolved and review metadata is written back.
Spec convergence relies on available cross-model feedback and requires frontmatter tagging to function properly. It cannot resolve open questions autonomously; human intervention is needed to address findings and update the spec until decision-completeness is achieved for approval.
maragudkCreate and update a living product spec at docs/spec.md.
kaptinlinAudit SPEC documents for completeness, consistency, and technical quality.
TNHoang2708Explore codebases in parallel using internal and external agents.
jikimeOrchestrate parallel AI subagents for multi-perspective code analysis and adversarial validation.
