UNPKG

@blundergoat/goat-flow

Version:

AI coding agent harness and local dashboard for Claude Code, OpenAI Codex, Google Antigravity, and GitHub Copilot - setup audits, guardrails, structured skills, deny hooks, and persistent learning loops.

52 lines 2.38 kB
/** * Context-free Markdown sections of the agent-setup quality prompt: the finding * rules, intentional-design notes, skill-testing protocol, template-integrity * checklist, rating rubrics/bands, and the closing reminder. Every helper here is * pure string assembly over a shared line buffer and needs no prompt context; * context-dependent sections stay in compose-quality-agent-setup.ts. */ /** * Append the per-finding Rules section (no tracked-file writes, evidence-based, * content-over-existence) that constrains every finding the reviewer reports. * * @param lines - prompt line buffer; appended to in place */ export declare function appendRules(lines: string[]): void; /** * Append the intentional-design notes the reviewer must NOT flag - gitignored * local state, the advisory `.active` pointer, unchecked task boxes, and the * lean post-ADR-014 config - so known-good shapes are not reported as findings. * * @param lines - prompt line buffer; appended to in place */ export declare function appendDesignNotes(lines: string[]): void; /** * Append Part 3 skill-testing instructions: the file-analysis vs live-invocation * options and the per-skill reporting-only probes, including the stop-and-report * rule when a probe attempts a tracked-file write or implementation. * * @param lines - prompt line buffer; appended to in place */ export declare function appendSkillTesting(lines: string[]): void; /** * Append the Part 6 skill-template integrity checklist: version-tag presence and * match, truncation/adaptation damage, and quick-vs-full depth coherence. * * @param lines - prompt line buffer; appended to in place */ export declare function appendSkillTemplateIntegrity(lines: string[]): void; /** * Append the Setup-Quality and System-Assessment prose prompts plus the two * 0-100 rating rubrics (four 0-25 axes each) the reviewer must score. * * @param lines - prompt line buffer; appended to in place */ export declare function appendRatingSections(lines: string[]): void; /** * Append the closing reminder: respond with the full prose assessment, write the * JSON report to the file path (not inline), and make no tracked-file edits. * * @param lines - prompt line buffer; appended to in place */ export declare function appendClosing(lines: string[]): void; //# sourceMappingURL=compose-quality-static-sections.d.ts.map