improvements-pipeline

Manage upstream service-improvement findings through filing, verification, and resolution workflows.

8|6|Updated Jul 31, 2026
One-click install
npx skills add https://github.com/stellar-experimental/stellar-raven --skill improvements-pipeline-stellar-experimental
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: improvements-pipeline
Source: https://github.com/stellar-experimental/stellar-raven/tree/main/.agents/skills/improvements-pipeline
Command: npx skills add https://github.com/stellar-experimental/stellar-raven --skill improvements-pipeline-stellar-experimental

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve? Tracking defects in upstream Stellar ecosystem services (docs, APIs, search) is error-prone: findings get duplicated, statuses drift from reality, upstream issues go stale, and resolved records linger. This Skill enforces a disciplined lifecycle for the improvements/ pipeline so every finding has reproducible evidence, a resolved intake owner, and a verifiable outcome. ## Core Features & Use Cases - Finding lifecycle management: File, dedupe, and transition findings through proposed, verified, reported-upstream, declined-upstream, and fixed-upstream statuses with evidence bars. - Standardized upstream filing: Generate owner-facing GitHub issues with a fixed five-section body, ASD-STE100-inspired writing style, and intake-based repo resolution via npm run improvements:file. - Follow-up and resolution loops: Re-run triggers and probes, classify outcomes (fixed, still-repro, closed-unfixed, superseded), and drain resolved records into improvements/resolved.json with reviewer verification. - Use Case: After an eval round reveals that Stellar docs search fails to surface the CLI install page, use this Skill to file a deduplicated finding, resolve the intake owner, draft and file the upstream issue, and later verify the fix with a live re-check before resolving the record. ## Quick Start Ask the agent to file a new upstream improvement finding for a Stellar service defect and regenerate the improvements index and lint checks.

Frequently Asked Questions about improvements-pipeline

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I file an upstream improvement finding for a Stellar service?

Create a finding file in the correct improvements/ collection with standard frontmatter plus Finding, Evidence, and Recommendation sections, then run npm run improvements:file with --dry-run to preview before posting. Dedupe against existing ids and recurrences first, and resolve the owner repo through improvements/intake.json.

What statuses can an improvement finding have?

Findings move through proposed, verified, reported-upstream, declined-upstream, and fixed-upstream. Each transition requires evidence: live re-execution for verified, a durable filed ref for reported-upstream, and a live re-check confirming the fix for fixed-upstream.

How do I check whether a reported upstream issue was fixed?

Inspect the GitHub issue or PR state with gh or GitHub MCP tools, then re-run the original trigger or probe rather than trusting the closed state. Only move the finding to fixed-upstream after observing the fix live, and record dated evidence plus the resolving ref.

Why does improvements lint fail on my finding?

Lint fails on malformed frontmatter, invalid status or service values, missing evidence or declined dispositions, stale generated INDEX.md, intake gaps, or evidence containing a GitHub URL while the status is still proposed or verified. Run npm run improvements:lint -- --live to also validate intake repos and recorded GitHub refs.

When should I comment on an open upstream issue?

Comment only for substantive maintainer activity, a claimed fix needing verification, materially new evidence or regressions, or author-owned PR work like requested changes. Silence is the default on untouched open issues; routine recurrences stay local in the finding record.

Can search-ranking problems in Stellar docs be fixed directly?

Yes, search-mechanism gaps in stellar-docs can be remediated with operator Algolia credentials, but only for general mechanisms proven on the read-only A/B harness first. Content gaps stay upstream on the docs repo, and per-query rules are prohibited as overfitting.