v1.0.0 Forge Console

Add a product-level Skill Forge console with doctor, demo, forge, evolve, install, uninstall, feedback, replay, release-check, and version commands. Keep the existing lower-level scripts as stable automation interfaces.

Promote the package to v1.0.0, document ClawHub-installed console usage, add Forge Console reference docs, and extend release checks with console doctor and demo smoke coverage.

Author: Skill Forge Maintainers

README.md

@@ -13,10 +13,19 @@ This repository contains the `skill-forge` OpenClaw-compatible skill. It also su
 
 `skill-forge` is a milestone self-improvement product for OpenClaw-style agents.
 
-Current version: `v0.5.0 "Signed Gate"`.
+Current version: `v1.0.0 "Forge Console"`.
 
 It turns repeated capability gaps into reviewed skill candidates. The product deliberately separates learning, generation, validation, and installation so an agent can improve itself without silently polluting its own skill set.
 
+`v1.0.0` introduces the Forge Console, a single product entry point over the full lifecycle:
+
+```bash
+python3 skill-forge/scripts/skill_forge.py doctor --json
+python3 skill-forge/scripts/skill_forge.py demo --json
+python3 skill-forge/scripts/skill_forge.py forge --output ./generated --install plan --json
+python3 skill-forge/scripts/skill_forge.py release-check --json
+```
+
 ## Quick Start
 
 ### 1. Install the skill into OpenClaw
@@ -37,6 +46,13 @@ The ClawHub package is published as `skills-forge` because `skill-forge` is alre
 
 ClawHub may show a safety warning because this skill includes Telegram approval calls and local install/rollback scripts. Review the source first; for non-interactive installs, pass `--force` only after review.
 
+After a ClawHub install, run the console from the installed package:
+
+```bash
+python3 ~/.openclaw/workspace/skills/skills-forge/scripts/skill_forge.py doctor --json
+python3 ~/.openclaw/workspace/skills/skills-forge/scripts/skill_forge.py demo --json
+```
+
 For local development, symlink this repository into OpenClaw:
 
 ```bash
@@ -63,26 +79,35 @@ Edit `~/.openclaw/skill-forge.env` and set your own Telegram bot token and chat
 ### 3. Run release checks
 
 ```bash
-python3 tests/run_release_checks.py
+python3 skill-forge/scripts/skill_forge.py release-check
 ```
 
-### 4. Generate a candidate skill
+### 4. Check readiness
 
 ```bash
-python3 skill-forge/scripts/forge_pipeline.py \
-  --source skill-forge/examples/sample-feature-requests.md \
-  --source skill-forge/examples/sample-errors.md \
-  --output ./generated \
-  --eval hidden \
-  --install plan \
-  --agent-name StudyAgent \
+python3 skill-forge/scripts/skill_forge.py doctor \
+  --env-file ~/.openclaw/skill-forge.env \
   --json
 ```
 
-### 5. Install only after Telegram approval
+### 5. Run the safe demo
 
 ```bash
-python3 skill-forge/scripts/forge_pipeline.py \
+python3 skill-forge/scripts/skill_forge.py demo --json
+```
+
+Expected behavior:
+
+- detects the bundled `citation-helper` example
+- generates a candidate skill
+- validates it as `grade=milestone`
+- prints a plan-only install proposal
+- does not mutate installed skills
+
+### 6. Install only after Telegram approval
+
+```bash
+python3 skill-forge/scripts/skill_forge.py forge \
   --source skill-forge/examples/sample-feature-requests.md \
   --source skill-forge/examples/sample-errors.md \
   --output ./generated \
@@ -107,6 +132,7 @@ Reply to the Telegram approval message with `同意安装` or `yes`.
 - Records later usage feedback and proposes reviewed skill updates instead of self-mutating installed skills.
 - Runs scheduled nightly health reviews during sleep hours and can propose safe update candidates.
 - Builds redacted replay cases from feedback and blocks update candidates that regress on replay evaluation.
+- Provides a unified Forge Console (`scripts/skill_forge.py`) for doctor checks, demos, generation, signed install plans, feedback capture, replay checks, and release gates.
 
 ## Repository Layout
 
@@ -125,6 +151,7 @@ skill-forge/
     │   ├── detect_skill_opportunities.py
     │   ├── evaluate_skill_candidate.py
     │   ├── forge_pipeline.py
+    │   ├── skill_forge.py
     │   ├── generate_skill_scaffold.py
     │   ├── validate_skill_candidate.py
     │   ├── install/
@@ -195,12 +222,7 @@ Root-level wrappers remain under `skill-forge/scripts/*.py` for backward compati
 Run the plan-only pipeline (no installs, no Telegram, no secrets needed):
 
 ```bash
-python3 skill-forge/scripts/forge_pipeline.py \
-  --source skill-forge/examples/sample-feature-requests.md \
-  --source skill-forge/examples/sample-errors.md \
-  --output ./generated-milestone \
-  --eval hidden \
-  --json
+python3 skill-forge/scripts/skill_forge.py demo --json
 ```
 
 Expected behavior:
@@ -220,7 +242,7 @@ export TELEGRAM_CHAT_ID="..."
 # When unset, the secret is derived deterministically from TELEGRAM_BOT_TOKEN.
 export SKILL_FORGE_APPROVAL_SECRET="..."
 
-python3 skill-forge/scripts/forge_pipeline.py \
+python3 skill-forge/scripts/skill_forge.py forge \
   --source skill-forge/examples/sample-feature-requests.md \
   --source skill-forge/examples/sample-errors.md \
   --output ./generated-milestone \
@@ -239,9 +261,7 @@ What `--install telegram` does internally:
 Dry-run preview without sending a Telegram message:
 
 ```bash
-python3 skill-forge/scripts/forge_pipeline.py \
-  --source skill-forge/examples/sample-feature-requests.md \
-  --source skill-forge/examples/sample-errors.md \
+python3 skill-forge/scripts/skill_forge.py forge \
   --output ./generated-milestone \
   --install telegram \
   --telegram-dry-run approve \
@@ -253,6 +273,23 @@ This issues a `mode=dry-run` token and exits with `install_status=dry-run-blocke
 
 ## Manual Workflow
 
+The Forge Console is the product-level interface. The lower-level scripts remain public and stable for automation.
+
+### Console Commands
+
+```bash
+python3 skill-forge/scripts/skill_forge.py version
+python3 skill-forge/scripts/skill_forge.py doctor --json
+python3 skill-forge/scripts/skill_forge.py demo --json
+python3 skill-forge/scripts/skill_forge.py forge --output ./generated --install plan --json
+python3 skill-forge/scripts/skill_forge.py evolve --skill citation-helper --output ./generated-updates --json
+python3 skill-forge/scripts/skill_forge.py install ./generated/citation-helper --json
+python3 skill-forge/scripts/skill_forge.py uninstall citation-helper --json
+python3 skill-forge/scripts/skill_forge.py feedback --skill citation-helper --rating negative --feedback "Needs stronger APA handling." --json
+python3 skill-forge/scripts/skill_forge.py replay --skill citation-helper --skill-dir ./generated-updates/citation-helper --json
+python3 skill-forge/scripts/skill_forge.py release-check --json
+```
+
 ### 1. Detect opportunities
 
 ```bash
@@ -597,6 +634,7 @@ This version is considered a milestone because it has a real closed loop:
 - `v0.4.2 "Telegram Gate"`: mandatory Telegram approval for install mutations
 - `v0.4.3 "Safety Tightening"`: safe skill slug containment for evolution updates and clearer registry safety posture
 - `v0.5.0 "Signed Gate"`: HMAC-signed approval tokens bound to skill identity and source content hash; missing or non-skill sources are refused before token issuance and before install mutation; dry-run approvals can no longer install or claim Telegram audit; replay regression gate compares candidate against installed baseline; manifest writes are atomic and `flock`-locked; redactor covers IPv4/IPv6, phone numbers, JWTs, common cloud/Git tokens, Telegram bot tokens, and PEM private keys
+- `v1.0.0 "Forge Console"`: product-level command surface with doctor, demo, forge, evolve, install, uninstall, feedback, replay, release-check, and version commands; release checks cover the console smoke path
 
 ## License
 

RELEASE.md

@@ -19,7 +19,7 @@ Manual checks:
 Release tag:
 
 ```bash
-git tag v0.5.0
+git tag v1.0.0
 ```
 
 Published package slugs:

skill-forge/SKILL.md

@@ -7,7 +7,7 @@ description: Detect repeated capability gaps, convert recurring user needs into
 
 Use this skill to turn repeated demand into a reviewed skill candidate.
 
-Current version: `v0.5.0 "Signed Gate"`.
+Current version: `v1.0.0 "Forge Console"`.
 
 ## Core jobs
 
@@ -20,6 +20,7 @@ Current version: `v0.5.0 "Signed Gate"`.
 7. Record feedback from later usage and propose reviewed updates instead of mutating installed skills directly.
 8. Provide the Somnia runtime used by scheduled nightly review to find skills with bugs, weak scores, or update-worthy feedback.
 9. Use redacted replay cases to check whether candidates actually cover real feedback-derived tasks.
+10. Provide a product console (`scripts/skill_forge.py`) so agents can run doctor checks, demos, forge flows, evolution, feedback capture, replay, and release gates from one entry point.
 
 ## Trigger Cues
 
@@ -47,6 +48,19 @@ Use this skill when the user or agent mentions:
 
 ## Commands
 
+### Forge Console
+
+Prefer the product console for normal use:
+
+```bash
+python3 {baseDir}/scripts/skill_forge.py doctor --json
+python3 {baseDir}/scripts/skill_forge.py demo --json
+python3 {baseDir}/scripts/skill_forge.py forge --output ./generated --install plan --json
+python3 {baseDir}/scripts/skill_forge.py release-check --json
+```
+
+Use lower-level scripts only when a workflow needs exact control over one stage.
+
 ### Detect
 
 Run:
@@ -288,13 +302,16 @@ References:
 - `references/heuristics.md`
 - `references/skill-quality-rubric.md`
 - `references/milestone-architecture.md`
+- `references/forge-console.md`
 
 Scripts:
 - `scripts/detect_skill_opportunities.py`
 - `scripts/evaluate_skill_candidate.py`
 - `scripts/generate_skill_scaffold.py`
 - `scripts/validate_skill_candidate.py`
 - `scripts/forge_pipeline.py`
+- `scripts/skill_forge.py`
+- `scripts/forge_console.py`
 - `scripts/install/propose_skill_install.py`
 - `scripts/install/telegram_approval.py`
 - `scripts/evolve/evolve_skill_pipeline.py`

skill-forge/VERSION

@@ -1 +1 @@
-0.5.0
+1.0.0

skill-forge/agents/openai.yaml

@@ -1,4 +1,4 @@
 interface:
   display_name: "Skill Forge"
-  short_description: "Detect, scaffold, validate, and signed-gate skill installs"
-  default_prompt: "Turn a repeated capability gap into a reviewed skill candidate. Detect from .learnings, scaffold by profile, validate and score, run hidden smoke evaluation, and require a Telegram-issued signed approval token (HMAC-bound to skill name, target, method, source path, and source content hash) before any install or update mutation."
+  short_description: "Forge Console for skill creation, validation, replay, and signed-gate installs"
+  default_prompt: "Use the Forge Console to turn repeated capability gaps into reviewed skill candidates. Start with doctor/demo when onboarding, then forge, validate, replay, and require a Telegram-issued signed approval token (HMAC-bound to skill name, target, method, source path, and source content hash) before any install or update mutation."

skill-forge/references/forge-console.md

@@ -0,0 +1,40 @@
+# Forge Console
+
+Forge Console is the product-level command surface for Skill Forge 1.0.
+
+Use it when users or agents need a complete lifecycle rather than a single internal script.
+
+## Command Map
+
+| Command | Purpose | Mutation risk |
+| --- | --- | --- |
+| `doctor` | Check local readiness: package, Python, OpenClaw path, Telegram config, approval secret, ClawHub, CI, git state | none |
+| `demo` | Run the bundled citation-helper example in plan-only mode | none |
+| `forge` | Detect, scaffold, validate, evaluate, and optionally request install approval | install only with signed approval |
+| `install` | Print or apply an install proposal for a candidate skill | apply requires signed approval |
+| `uninstall` | Print or apply uninstall / rollback for an installed skill | apply requires signed approval |
+| `feedback` | Record redacted feedback for later evolution | writes feedback JSONL |
+| `evolve` | Propose a reviewed update from feedback and replay checks | install only with signed approval |
+| `replay` | Score a skill directory against replay cases | none |
+| `release-check` | Run release gates and secret scan | none |
+| `version` | Print product version | none |
+
+## Product Defaults
+
+- Default install mode is plan-only.
+- Telegram approval is required for any install or uninstall mutation.
+- Dry-run approval tokens cannot mutate through the pipeline.
+- `doctor` should be the first command when onboarding a new machine.
+- `demo` should be the first command when validating the installation.
+- `release-check` should pass before GitHub tags, ClawHub publishing, or public demos.
+
+## Recommended User Flow
+
+```bash
+python3 scripts/skill_forge.py doctor --json
+python3 scripts/skill_forge.py demo --json
+python3 scripts/skill_forge.py forge --output ./generated --install plan --json
+python3 scripts/skill_forge.py release-check --json
+```
+
+Only move to `--install telegram` once the plan-only candidate is reviewed.

skill-forge/scripts/forge_console.py

@@ -0,0 +1,7 @@
+#!/usr/bin/env python3
+import runpy
+from pathlib import Path
+
+
+if __name__ == "__main__":
+    runpy.run_path(str(Path(__file__).resolve().parent / "skill_forge.py"), run_name="__main__")