Releases: magnus919/agent-skills
Releases · magnus919/agent-skills
Release list
v0.5.0
0.5.0 (2026-07-27)
Features
- add evidence-aware life coaching skill (#146) (795d142)
- add Gibson cyberpunk skill (#149) (daece12)
- add harness adapter contract and reproducible eval run artifacts (#130) (e83558a)
- add llama.cpp operations skill (#145) (58457b4)
- add release-grade repeated and rubric-based skill evaluation (#136) (0a80bda), closes #106
- expand Raleigh civic data skill (eff14ca)
- grafana: add operations skill (4c824b7)
- grafana: add operations skill (e304366)
- raleigh: add fire protection proximity lookup (#144) (1f5cd6a)
- raleigh: add first-class RFD incident workflows and normalize the 2026 schema (#140) (0ca5c24), closes #122
- raleigh: add first-class RPD incident workflows (#139) (50be18b), closes #121
- raleigh: add guarded fire report lookups (12b650e)
- raleigh: add guarded fire report lookups (4eaf5aa)
- raleigh: add official public safety statistics (b030f26)
- raleigh: add Raleigh-Wake ECC active incident feed adapter (#142) (d4ab656)
- raleigh: expand civic data skill (903ea05)
- raleigh: expose official public safety statistics (dcd6f59)
- run isolated paired candidate and baseline skill evaluations (#133) (390f3e3)
- skill: add PACE plan lifecycle (e62152c)
- skill: add PACE plan lifecycle (64383d9)
- validate eval manifest coverage states (#129) (a617cca)
Bug Fixes
- align eval artifact schemas with runtime (f49e670)
- allow incidental text in cyberpunk image briefs (6b467df)
- complete eval ratchet verification (#109) (1e10a56)
- enforce eval coverage ratchet for complete changed skill directories (#107) (fc8a895), closes #102
- harden paired evaluation isolation (cb1579f)
- life-coach: simplify capability onboarding (#148) (fdb00b4)
- llama-cpp: harden operational diagnostics (#147) (86ac280)
- raleigh: bound publication source contracts (cc26b4c)
- raleigh: enforce publication request contracts (1bd7b06)
- raleigh: fail on missing statistics sections (8becc22)
- raleigh: normalize report probe failures (f3dfb68)
- raleigh: normalize statistics source failures (07c2ae2)
- raleigh: reject malformed content identifiers (5f74b4d)
- raleigh: restore live RPD queries (cfbcfb0)
- raleigh: restore live RPD queries (4a23562)
- raleigh: validate included resource identifiers (9f2e4b6)
- raleigh: validate statistics response shapes (bf69aa1)
- reject duplicate eval fixture paths (cbe1b69)
- require causal layer maps for cyberpunk image briefs (34fab96)
- require causal layer maps for cyberpunk image briefs (34fab96)
- require causal layer maps for cyberpunk image briefs (c514c85)
- require evidence for Grafana dashboard defaults (4b60a2e)
- require evidence for Grafana dashboard defaults (4ecc536)
- skill: finalize PACE delivery evidence (f62a5de)
- skills: rename gibson-cyberpunk to cyberpunk (#150) (1005462)
- verification: preserve requested source fidelity (#127) (547aed2)
v0.4.0
0.4.0 (2026-07-22)
Features
- add Claude Code plugin marketplace (metadata-only catalog) (#80) (3c8b6cd), closes #76
- add Codex plugin packaging (single-plugin, metadata-only) (#83) (0f2aaf1)
- add generated llms.txt skill index (#93) (cde4a67), closes #78
- add neckbeard, an evidence-driven SDLC skill bundle (#81) (e167718)
- add programming-principles skill (14 classic software books) (#92) (1037324)
- eval coverage ratchet — gate new skills, track coverage, ratchet thresholds (#99) (be0c8df), closes #90
- qa-methodology: add ci-failure-triage and test-debugging references (#94) (08fafac)
- validate changed skill descriptions (#97) (c092a14)
Bug Fixes
v0.3.0
What's Changed
234 commits, 49 merged PRs since v0.2.2. The catalog grew from ~55 to 105 validated skills.
New Skills (37 added)
Engineering & Operations
- backend-engineering — API patterns, service architecture, DB access, error handling
- frontend-engineering — component architecture, state management, performance, responsive layout
- data-engineering — ETL/dbt, SQL analytical patterns, vector/graph/time-series ops
- ml-engineering — fine-tuning, evaluation, quantization, training infrastructure
- platform-engineering — IaC, CI/CD, containers, networking, secrets, observability
- qa-methodology — test strategy, automation, regression, data mgmt, perf/security testing
- esp32-development — build, flash, test, debug, and recover ESP32 firmware
- remote-systems-administration — SSH, Ansible, multi-host operations
- site-reliability-engineering — SLOs, incident command, error budgets
- secure-software-engineering — threat modeling, secure coding patterns
- api-design-and-evolution — API contracts, versioning, deprecation
- web-accessibility — WCAG compliance, audit patterns
- supabase — full Supabase skill (managed + self-hosted)
- vercel-eve — Vercel Eve platform skill
Methodology & Process
- verification-methodology — evidence-based acceptance criteria
- research-methodology — plan, conduct, evaluate, synthesize research
- security-audit-methodology — threat modeling, authorized reviews
- technical-documentation — create and review technical docs
- artifact-pyramids — organize durable research outputs
- technology-radar — build and maintain tech radars
- de-spin — claim-audit skill for truth/falsehood assessment
Architecture & Diagramming
- adr-authoring — architecture decision records
- c4-diagramming — C4 software architecture diagrams
- mermaid-diagrams — author, render, troubleshoot Mermaid
Product & Strategy
- product-methodology — product definition, PRDs, requirements
- product-design-and-ux — UX design methodology
- product-strategy — CPO methodology (North Star, RICE, Porter)
- strategy-frameworks — competitive and business strategy
- financial-modeling — financial models and analysis
- chief-of-staff-methodology — gatekeeping, force multiplication
- go-to-market — CMO methodology (Dunford, CAC/LTV, PLG/SLG)
- legal-strategy — CLO methodology (GDPR, IP, contracts, privacy)
- operational-design — COO methodology (VSM, scaling, SOC 2, RFP)
- org-design — CHRO methodology (Team Topologies, comp, culture)
Tools & Integrations
- lastfm — Last.fm music data API with discovery pipeline
- fireflies — Fireflies.ai meeting transcript queries
- linear — Linear project/issue management CLI
- jellyfin-cli — read-only media server navigation
- restic — recovery-first backup operations
- agent-evals-and-observability — framework-neutral agent evals
- daily-life-discovery — consent-based discovery conversations
- opensource-contributions — comprehensive OSS contribution guide
Bundles
- tailscale — 7 sub-skills for self-hosted Tailscale/Headscale (first bundle)
- workflow-architect — workflow discovery + skills bundle generator (3 sub-skills)
Skill Improvements
- nous-branding: style lane system, reference catalog, pitfalls doc, post-processing script
- kubernetes: SkillOpt hygiene cleanup, endpoint verification hardening
- meshcore-packet-capture: SkillOpt Epoch 1 optimization
- forgejo-cli: added
pr createsubcommand, fixed --labels integer passing - verification-methodology: accessible evidence collection
- supabase: hardened operational safety guidance
- esp32-development: evidence gate for hardware claims
Infrastructure & Governance
- CI: Ruby validator enforces SKILL.md format, README sections, catalog completeness
- CI: Python artifact checker validates tracked files
- CI: Python wheel builds for packaged skills
- Root README: enforced alphabetical catalog order and complete coverage
- Governance: aligned with shipped artifact types
- Added graphify-out/ to .gitignore
- Acknowledged plntrprotocol/nous-branding parallel development
Fixes
- Linear raw GraphQL mutations classified safely
- bundle-builder registers generated sub-skills with Hermes skill system
- research skill requires durable artifact preservation
- de-spin gates publication on completed checks
- Removed invalid Hermes skill metadata from ported skills
Full Changelog: v0.2.2...v0.3.0