production_agent_skill_lifecycle
Guides agents through designing, reviewing, and maintaining production-grade agent skills with trigger-focused frontmatter, behavioral workflows, anti-rationalization, and evidence gates. Use when creating or upgrading SEOSONA skills, importing external skill packs, or auditing whether a skill changes agent behavior under pressure.
- License 许可
- MIT
- 最近更新
- 2026-10-09
- Stars
- ★ 2
- Forks
- 1
01 前言 Front Matter
- name
- production_agent_skill_lifecycle
- description
- Guides agents through designing, reviewing, and maintaining production-grade agent skills with trigger-focused frontmatter, behavioral workflows, anti-rationalization, and evidence gates. Use when creating or upgrading SEOSONA skills, importing external skill packs, or auditing whether a skill changes agent behavior under pressure.
03 SKILL.md 正文
Production Agent Skill Lifecycle
Overview
Use this skill when a SEOSONA capability should become a reusable agent skill rather than a one-off note. The goal is to encode a repeatable workflow that agents can discover, execute, verify, and improve.
When To Use
- Creating a new skill from external research.
- Refactoring a vague knowledge note into an operational workflow.
- Auditing a skill that reads like documentation but does not change behavior.
- Importing skill packs from another agent ecosystem.
Do not use this for simple static reference notes. Use 2_KNOWLEDGE/raw_data/ for passive knowledge that has no repeatable workflow.
Workflow
-
Define the behavioral trigger.
- The frontmatter description must say what the skill does and when to use it.
- Avoid process summaries in the description; the agent must still read the full skill.
-
Write the operating loop.
- Use phases or numbered steps.
- Make every step observable.
- Name stop conditions and handoff boundaries.
-
Add anti-rationalization.
- List excuses that agents use to skip work.
- Pair each excuse with the operational reason it is unsafe.
-
Add red flags.
- Describe visible signs of drift, such as skipping verification, inventing source facts, or expanding scope without a plan update.
-
Add verification.
- Require command output, source links, screenshots, tests, or generated artifacts.
- "Looks good" is not evidence.
-
Keep progressive disclosure.
- Keep
SKILL.mdfocused. - Move long reference material into
references/. - Add scripts only when they are actually run by the workflow.
- Keep
-
Register and validate.
- Rebuild the capability graph/router after adding skills.
- Run the smallest relevant validation gate.
Skill Anatomy
Required:
SKILL.md- YAML frontmatter with
nameanddescription - A clear workflow and verification section
Recommended:
When To UseWorkflowCommon RationalizationsRed FlagsVerificationReferences
Optional:
references/for long checklists or examplesscripts/for repeatable helper commandstemplates/for generated artifacts
Common Rationalizations
| Rationalization | Reality |
|---|---|
| "This is obvious; the skill can be short." | Obvious steps are the first steps agents skip under pressure. |
| "The README explains it." | A README informs; a skill changes execution behavior. |
| "Verification can be done later." | A skill without exit evidence is a suggestion, not an operating contract. |
| "More context is always better." | Overloaded skills reduce discoverability and increase drift. |
Red Flags
- The description says only a topic, not a trigger.
- The body is mostly background prose.
- There is no exit checklist.
- The workflow asks the user for permission when local context can answer safely.
- The skill duplicates another skill instead of referencing it.
Verification
After creating or upgrading a skill, confirm:
- [ ] The description includes both what and when.
- [ ] The workflow has concrete steps.
- [ ] The skill has evidence-based verification.
- [ ] No absolute local paths were written.
- [ ] Capability validation or router rebuild was run.
04 同仓库更多 Skills
web_crawler_katana
High-speed web crawling using Katana patterns for deep site mapping, URL discovery, and content extraction.
server_log_analyzer
Analyzes Nginx/Apache server logs to identify Googlebot crawl patterns, errors, and performance bottlenecks.
schema_markup_generator
Auto-generates JSON-LD structured data for any page type (Article, Product, FAQ, LocalBusiness, etc.).
internal_linking_optimizer
Analyzes and optimizes the internal link graph of a website for SEO authority distribution.
google_trends_analyzer
Detects trending topics and seasonal patterns using Google Trends data for content planning and keyword strategy.
brand_context_analyzer
Tracks brand mentions, sentiment, and online reputation across web sources.