Agent Skills Pipeline v1.0

8-stage pipeline: Scout → Filter → Reader → Extractor → Score → Generator → Reviewer → Publisher - Scout: GitHub search with token auth + rate limit retry - Filter: Deterministic rules (language, stars, age, keywords) - Reader: Incremental context loading (README → docs → examples → code) - Extractor: LLM workflow extraction with JSON retry - Score: Rule-based evaluation (no LLM) - Generator: Standardized Hermes Skill format - Reviewer: Independent LLM review (separate from generator) - Publisher: Branch + PR to Gitea First run: 5 repos discovered, 0 extracted (correct — all frameworks, no workflows)
2026-08-05 05:51:06 +00:00
commit 8da8d703da
25 changed files with 1159 additions and 0 deletions
@@ -0,0 +1,103 @@
+"""Stage 6: Skill Generator — Transform workflow into standardized Hermes Skill."""
+import yaml
+import json
+
+def generate_skill(score_result, config):
+    """
+    Generate a standardized Hermes Skill package from an approved workflow.
+    """
+    if score_result.get("decision") != "PASS":
+        return {
+            "status": "BLOCKED",
+            "reason": "Score check failed",
+        }
+
+    workflow = score_result.get("workflow", {})
+    skill_name = workflow.get("skill_name", "unknown").lower().replace(" ", "-").replace("_", "-")
+    repo = score_result.get("repository", "")
+
+    # Generate SKILL.md (frontmatter + markdown body)
+    frontmatter = {
+        "name": skill_name,
+        "version": "1.0.0",
+        "description": workflow.get("goal", ""),
+        "inputs": workflow.get("inputs", []),
+        "steps": workflow.get("steps", []),
+        "outputs": workflow.get("outputs", []),
+        "tags": [],
+        "metadata": {
+            "source_repo": repo,
+            "extracted_at": "",
+            "confidence": workflow.get("confidence", 0),
+        },
+    }
+
+    skill_md = "---\n"
+    skill_md += yaml.dump(frontmatter, default_flow_style=False, sort_keys=False)
+    skill_md += "---\n\n"
+    skill_md += f"# {skill_name}\n\n"
+    skill_md += f"{workflow.get('goal', '')}\n\n"
+    skill_md += f"## Steps\n\n"
+    for i, step in enumerate(workflow.get("steps", []), 1):
+        skill_md += f"{i}. {step}\n"
+    skill_md += f"\n## Inputs\n\n"
+    for inp in workflow.get("inputs", []):
+        skill_md += f"- {inp}\n"
+    skill_md += f"\n## Outputs\n\n"
+    for out in workflow.get("outputs", []):
+        skill_md += f"- {out}\n"
+    skill_md += f"\n## Failure Modes\n\n"
+    for fm in workflow.get("failure_modes", []):
+        skill_md += f"- {fm}\n"
+    skill_md += f"\n## Source\n\n"
+    skill_md += f"Extracted from: [{repo}]({repo})\n"
+    skill_md += f"Confidence: {workflow.get('confidence', 0)}\n"
+
+    # Generate examples.md
+    examples_md = f"# Examples: {skill_name}\n\n"
+    examples_md += f"## Usage Example\n\n"
+    examples_md += f"```python\n# How to use this skill\n# Inputs: {', '.join(workflow.get('inputs', []))}\n# Process: {' → '.join(workflow.get('steps', [])[:3])}\n# Outputs: {', '.join(workflow.get('outputs', []))}\n```\n"
+
+    # Generate commands.md
+    commands_md = f"# Commands: {skill_name}\n\n"
+    commands_md += f"## Available Commands\n\n"
+    commands_md += f"- `/skill {skill_name}` — Load this skill\n"
+    commands_md += f"- `/run {skill_name}` — Execute workflow\n"
+
+    # Generate metadata.json
+    metadata = {
+        "name": skill_name,
+        "version": "1.0.0",
+        "goal": workflow.get("goal", ""),
+        "inputs": workflow.get("inputs", []),
+        "steps": workflow.get("steps", []),
+        "outputs": workflow.get("outputs", []),
+        "failure_modes": workflow.get("failure_modes", []),
+        "confidence": workflow.get("confidence", 0),
+        "explanation": workflow.get("explanation", ""),
+        "source_repo": repo,
+        "score": score_result.get("score", 0),
+    }
+
+    # Generate tests.md
+    tests_md = f"# Tests: {skill_name}\n\n"
+    tests_md += f"## Test Checklist\n\n"
+    tests_md += f"- [ ] Workflow has at least 3 steps\n"
+    tests_md += f"- [ ] All inputs are defined\n"
+    tests_md += f"- [ ] All outputs are defined\n"
+    tests_md += f"- [ ] Failure modes are documented\n"
+    tests_md += f"- [ ] Skill can be loaded without errors\n"
+
+    return {
+        "status": "GENERATED",
+        "skill_name": skill_name,
+        "repository": repo,
+        "files": {
+            "SKILL.md": skill_md,
+            "examples.md": examples_md,
+            "commands.md": commands_md,
+            "metadata.json": json.dumps(metadata, indent=2),
+            "tests.md": tests_md,
+        },
+        "metadata": metadata,
+    }