| 2026-10-15 10:31 |
v3.7.0 — Added KGM registration step to Post-Creation Checklist — agent registers skill entity in KGM after creation so hey signal dispatch can find it |
| 2026-10-02 22:29 |
v2.1.0 — P1: added weak-model compatibility checklist (formatting directives, concrete examples, context-aware references, wrapper scripts). P7: added script composition safety (SIGPIPE, set -euo pipefail patterns). |
| 2026-09-24 16:05 |
v3.6.0 — Add corollary to Code-First principle: lower cognitive bar for model callers — zero-arg wrappers, auto-detect context, avoid similar script names |
| 2026-09-10 13:56 |
vv3.5.0 — Reversed default scope from USER to PROJECT. Scope governs discovery and context loading, not operational target. Added criteria table for scope decision. PROJECT avoids signal phrase bloat in unrelated sessions; promote to USER when cross-project demand is proven. |
| 2026-09-10 12:59 |
vv3.4.0 — Added Principle 6 (Context Economy): soft 300-line threshold with two-step audit, references/ directory conventions, standard reference file names. Added Principle 7 (Error Contract): semantic exit codes (0/1/2/3), privilege gate pattern (exit 3), script diagnostic output conventions, SKILL.md error handling obligations (idempotency table, recovery paths, troubleshooting table), prescribed agent behavior per exit code. Made Skill Check mandatory in Post-Creation Checklist after every creation/update. Expanded Business Process Modeling with privilege gate markup pattern. Updated skill-schema.md v2.3.0: added references/ directory convention, exit code convention, standard file names. |
| 2026-09-04 15:01 |
v3.3.0 — Added Script Comment Separation of Concerns principle and cross-skill execution priority order to SKILL.md Step 5. Added Script Comment Rules section to skill-schema.md (v2.2.0). |
| 2026-08-26 22:24 |
v— — Terminology: mode → scope. Added LITE scope guard note where applicable. |
| 2026-08-20 10:30 |
v3.2.0 — Added explicit ⛔ GATE to Post-Creation Checklist requiring user permission before skipping Principle 4 (Spec/Tests) or Principle 5 (Security) |
| 2026-08-13 19:22 |
v3.1.0 — Externalized changelogs: SKILL.md template now scaffolds CHANGELOG.md as separate file with reference link in SKILL.md; updated Post-Creation Checklist, Skill Check P3, and edit-existing guidance; updated references/skill-schema.md to v2.1.0 with new Changelog Rules |
| 2026-08-13 12:30 |
v3.0.0 — Breaking: description field redefined as signal phrases (Command: verb+noun(s), Query: noun(s)); removed metadata.signals from schema; added Signal Phrase Rules with three quality principles (Concise, No redundant, Complete) and smell test; added Opening Paragraph requirement; updated SKILL.md template, Post-Creation Checklist (Step 6), and Skill Check P1; updated references/skill-schema.md to v2.0.0 |
| 2026-08-10 16:11 |
v2.1.1 — Strengthened Gotchas from suggestion (“Consider”) to requirement (“Include”); added Gotchas section to SKILL.md template; added P1 checklist item for Gotchas/Known Issues section |
| 2026-08-10 15:58 |
v2.1.0 — P1: added inline code boundary guidance — reference code (templates, examples, one-liners, manual fallbacks) belongs in SKILL.md; execution code (multi-step, verification, discovery, conditionals) belongs in scripts/; added checklist item for not flagging reference code as 🟡 |
| 2026-08-10 12:01 |
v2.0.0 — Added Principle 5 (Security & Trust Boundary) to Skill Check: scripts scope, secrets handling, network calls, prompt injection; “Four Principles” → “Five Principles”; added “Real Expertise” anti-pattern guidance and Gotchas pattern to Step 1 (Capture Intent); added “loose steps → instructions, fragile steps → code” aphorism to Agent-as-Orchestrator |
| 2026-08-10 11:18 |
v1.9.0 — Strengthened P1 to “Accuracy, Consistency & Testability (Code-First)”: skills are prescriptive SOPs verified by execution; code is primary vehicle for accuracy, consistency, conciseness, and testability; added could-be-script 🟡 flag and knowledge-extraction flag to checklist; moved Concise from P3 to P1; P3 renamed to “Traceable & Well-Formatted” |
| 2026-08-08 13:39 |
v1.8.0 — Skill Check expanded to 4 principles: added Principle 4 (Verifiable Specification & Test) with spec table and testcase mapping examples; Principle 1 now flags inline-only operations as critical |
| 2026-08-08 13:20 |
v1.7.0 — Added Skill Check mode: audit skills against 3 principles (Accuracy/Consistency, Autonomous/Currency, Concise/Traceable); checklist-driven procedure with severity-rated findings report |
| 2026-08-08 10:55 |
v1.6.0 — Added “Defensive file templates” writing guidance: skills that write external files must include exact templates with warning blocks; added writes-files optional field to skill-schema.md |
| 2026-08-04 23:47 |
v1.5.0 — Added canonical SKILL.md frontmatter schema (references/skill-schema.md); template now uses quoted 3-part semver ("1.0.0"); post-creation checklist requires metadata.version; writing guidance references schema doc |
| 2026-07-27 18:35 |
v1.4 — Added metadata.signals to SKILL.md template and post-creation checklist; added signal quality check; updated description guidance to emphasize conciseness (loaded every session via built-in skills listing) |
| 2026-07-14 14:51 |
v1.3 — Added “Name consistency” check to post-creation checklist: name field must match directory name per Agent Skills open standard (agentskills.io/specification) |
| 2026-07-13 16:11 |
v1.2 — Added “Skill Design Principles” section: non-interactive scripts, agent-as-orchestrator pattern, business process modeling |
| 2026-07-13 11:19 |
v1.1 — Renamed from skill-creator to skill-gen for naming consistency with okf-bundle-gen, bash-completion-gen; updated all internal path references |
| 2026-07-10 17:05 |
v1.0 — Initial proxy skill: simple + advanced modes, upstream Anthropic skill-creator integration, AgentFS post-creation checklist, agent compatibility notes |