jbsanf/skillver

skillver

Recover and maintain repo-native feature memory during feature work, and preserve verified reusable techniques after substantive investigation.

View source

How to use

  1. Copy the installation command and review the source package before installing it in your own agent environment.
  2. Check the source document for required tools, credentials and supported hosts.
  3. Describe your input and desired deliverable; review the output against the source material.
npx skills add https://github.com/jbsanf/skillver --skill skillver

Before you start

  • Installation does not include a model subscription, API allowance or permission to access your files.
  • Repository stars and installation counts describe activity, not an independent quality or security test.
Original skill document

Rendered from the source repository. Headings, examples, code, tables, links, and referenced images are preserved.

Skillver

Keep project state in .agent-memory/features/ and reusable procedures in .agents/skills/. A procedure may depend on project-specific conventions; current product status belongs in a feature track.

Use repository Markdown and existing documentation. Do not introduce external memory stores.

Scope and safety

Follow the user's current scope and the host's execution rules. During authorized implementation, maintain relevant local memory and extract qualifying skills. Queries, retrospectives, reviews, and planning-only requests produce answers or proposals unless writing is also authorized.

Never persist credentials, tokens, internal URLs, or customer data in tracks, skills, examples, or verification evidence.

Keep memory version-controlled under the project's workflow. This does not authorize commits or changes to repository policy.

Feature memory

For non-trivial feature work, read the existing .agent-memory/features/README.md and the relevant track. Follow its source links and verify claims that affect the task against current evidence. Distinguish implemented behavior from approved decisions and future plans; report discrepancies.

Create a missing track when the authorized work establishes durable feature knowledge. For setup, start with requested or clearly documented features and expand as needed.

Before creating or changing memory, read references/feature-memory.md. Use the index template and the track template for new files.

Preserve existing documentation locations, headings, and lifecycle conventions. Summarize and link rather than duplicate or relocate documents.

Record meaningful changes to behavior, decisions, risks, verification, or authoritative sources. Update the track, dated changelog, and affected index fields together. Do not rewrite unchanged memory.

Validate touched tracks and their index entries. Audit the whole memory during setup or an explicit documentation check; expand validation when findings justify it. Report unresolved drift.

Skill extraction

After substantive investigation, consider whether the session produced reusable, specific, non-trivial, verified knowledge. Zero extractions is a normal outcome.

Before creating or changing a skill, read references/skill-extraction.md. Search available related skills and prefer a useful refinement over duplication. Finding a personal or third-party skill does not authorize editing it.

Use the extraction template. Record applicability, the verified procedure, relevant constraints, and observed verification. Preserve failed approaches when they prevent a likely mistake.

Do not promote hypotheses or copies of documentation into verified procedures. Documented techniques qualify when the skill adds a proven contextual choice, sequence, or constraint.

A retrospective presents qualifying candidates and their rationale. An explicit request to save a skill authorizes local extraction, subject to the same evidence and safety requirements.

Replace template placeholders. Optional sections may be omitted; feature tracks follow the section rules in the memory reference.

Related by task