Skip to main content
Skill Library changelog

What changed in the public Skill Library.

Release notes for public workflow libraries, filters, Governed AI Workflow examples, guardrail pages, and field-guide updates.

Short answer

Why keep a Skill Library changelog?

A Skill Library should improve as workflows fail, tools change, and review rules get sharper. The changelog shows what changed so teams can decide what to inspect or adapt.

Browse the Skill Library
2026-08-01

Skill Library portfolio rewritten

Replaced the repeated five-skill packaging pattern with focused, research-specific skills whose size and structure follow the workflow instead of a shared template.

  • Rewrote all 35 workflow libraries into 54 focused Agent Skills with trigger-specific descriptions and research-derived methods.
  • Regenerated every library, individual-skill, and complete-portfolio ZIP from the new authored sources.
  • Expanded the authored evaluation suite to 181 domain-specific scenarios with application, variation, gap, counterexample, and authority-boundary coverage.
  • Marked authored evaluation hypotheses as not executed so planned tests cannot be mistaken for completed behavioral evidence.
2026-07-19

Agent tool permission lease review added

Published an agent tool permission lease workflow library for making temporary tool authority task-bound, time-boxed, revocable, and testable against stale replay.

  • Added Agent Tool Permission Lease Review to the public Skill Library catalog.
  • Added five downloadable Agent Skills for lease records, expiry and closure, renewal routing, stale replay tests, and revocation audit gates.
  • Added eval scenarios for clean, messy, sensitive, unsupported-commitment, and prompt-injection inputs.
  • Added public ZIP downloads for the library and individual skills.
2026-07-18

Negative authorization test review added

Published a negative authorization workflow library for proving denied agent tool paths fail before real write, shell, browser, network, MCP, memory, or communication authority is granted.

  • Added Negative Authorization Test Review to the public Skill Library catalog.
  • Added five downloadable Agent Skills for denied-path matrices, policy fixtures, prompt-injection denial cases, non-effect checks, and authorization regression gates.
  • Added eval scenarios for clean, messy, sensitive, unsupported-commitment, and prompt-injection inputs.
  • Added public ZIP downloads for the library and individual skills.
2026-07-17

Agent interruption budget review added

Published an agent interruption budget workflow library for deciding when agents should interrupt now, batch review, stay in shadow mode, or stop before reviewer capacity becomes the failure mode.

  • Added Agent Interruption Budget Review to the public Skill Library catalog.
  • Added five downloadable Agent Skills for interrupt routing, batch-review queues, shadow-mode lanes, stop and capacity gates, and review-signal learning loops.
  • Added eval scenarios for clean, messy, sensitive, unsupported-commitment, and prompt-injection inputs.
  • Added public ZIP downloads for the library and individual skills.
2026-07-16

Handoff ownership transfer review added

Published a handoff ownership workflow library for moving work between agents, humans, orchestrators, and services without losing responsibility, evidence, artifact versions, or authority limits.

  • Added Handoff Ownership Transfer Review to the public Skill Library catalog.
  • Added five downloadable Agent Skills for packet writing, receiver synthesis, provenance and version tracing, authority transfer, and blocked handoff repair.
  • Added eval scenarios for clean, messy, sensitive, unsupported-commitment, and prompt-injection inputs.
  • Added public ZIP downloads for the library and individual skills.
2026-07-15

Agent escalation trigger review added

Published an agent escalation workflow library for turning risky exception paths into stop, ask, downgrade, escalate, fail-closed, and owner-reviewed resume decisions.

  • Added Agent Escalation Trigger Review to the public Skill Library catalog.
  • Added five downloadable Agent Skills for trigger intake, action routing, structured clarification, fail-closed evidence packets, and resume decisions.
  • Added eval scenarios for clean, messy, sensitive, unsupported-commitment, and prompt-injection inputs.
  • Added public ZIP downloads for the library and individual skills.
2026-07-09

Prompt mismatch log review added

Published a prompt mismatch workflow library for recording intent, prompt version, rendered context, downstream result, failure labels, route decisions, and regression cases before prompt rewrites.

  • Added Prompt Mismatch Log Review to the public Skill Library catalog.
  • Added five downloadable Agent Skills for mismatch envelopes, failure labels, clarification routing, regression case promotion, and prompt-change decisions.
  • Added eval scenarios for clean, messy, sensitive, unsupported-commitment, and prompt-injection inputs.
  • Added public ZIP downloads for the library and individual skills.
2026-07-08

Skill maintenance schedule added

Published a Skill maintenance workflow library for owner metadata, dependency drift, event-triggered reviews, lifecycle decisions, and retirement or rollback planning.

  • Added Skill maintenance schedule to the public Skill Library catalog.
  • Added five downloadable Agent Skills for maintenance records, dependency drift, review scheduling, lifecycle decisions, and retirement or rollback planning.
  • Added eval scenarios for clean, messy, sensitive, unsupported-commitment, and prompt-injection inputs.
  • Added public ZIP downloads for the library and individual skills.
2026-07-06

Confidence calibration review added

Published a confidence calibration workflow library for separating confidence signals from permission before agents get more authority.

  • Added Confidence calibration review to the public Skill Library catalog.
  • Added five downloadable Agent Skills for action authority, confidence recording, baseline comparison, abstention routing, and override logging.
  • Added eval scenarios for clean, messy, sensitive, unsupported-commitment, and prompt-injection inputs.
  • Added public ZIP downloads for the library and individual skills.
2026-06-29

AI adoption workflow redesign added

Published an AI adoption workflow redesign library for diagnosing adoption plateaus, reframing usage dashboards, selecting one measurable workflow, and piloting with review gates.

  • Added AI adoption workflow redesign to the public Skill Library catalog.
  • Added five downloadable Agent Skills for adoption diagnosis, workflow selection, metric reframing, workflow embedding, and pilot planning.
  • Added eval scenarios for clean, messy, sensitive, unsupported-commitment, and prompt-injection inputs.
  • Added public ZIP downloads for the library and individual skills.
2026-06-26

Agent launch evidence review added

Published an agent launch evidence workflow library for accountability, tool permission manifests, disagreement logs, runnable examples, and launch gates.

  • Added Agent launch evidence review to the public Skill Library catalog.
  • Added five downloadable Agent Skills for launch evidence review.
  • Added eval scenarios for clean, messy, sensitive, unsupported-commitment, and prompt-injection inputs.
  • Added public ZIP downloads for the library and individual skills.
2026-06-22

Agent governance libraries added

Published three agent-governance workflow libraries from the VibeSec Advisory Skill Library repository onto the public website.

  • Added Agent workflow control review to the public Skill Library catalog.
  • Added Agent tool authority review to the public Skill Library catalog.
  • Added AI workflow eval and lifecycle review to the public Skill Library catalog.
  • Added downloadable library ZIPs and individual skill ZIPs for each new library.
2026-05-29

Remaining marketing backlog implementation

Expanded the public Skill Library path with filters, topic hubs, comparison content, fictional workflow examples, and clearer Manual fit checks.

  • Added source-data-driven role and workflow filters to the Skill Library index.
  • Added CollectionPage and ItemList structured data to the Skill Library index.
  • Added Manual fit-check copy to every public Skill Library detail page.
  • Added topic hubs for GTM guardrails, Skill Libraries, AI agent security, and FORGE methodology.
  • Added a public prompt dumps versus Skill Libraries comparison page.
  • Added fictional and sanitized workflow examples for operators evaluating public libraries.
2026-05-29

Public Skill Library proof path

Aligned the Skill Library with the public field-guide path: inspect a public library, compare it to workflow examples, then adapt it carefully before any agent gets real tool or data access.

  • Kept public library downloads free and inspectable.
  • Retired commercial bridge language from the public changelog while keeping the free libraries inspectable and downloadable.
  • Kept Stripe, checkout, payment logic, and live product IDs unchanged.
Keep going

Use the changelog as a review queue.

When a workflow library changes, review the assumptions before reusing it in an agent or team process. New releases are free starting points, not proof that a workflow is safe in your environment.

Review workflow examples