Hermes Ponytail - Setup Guide¶
Ponytail makes your Hermes agent think like the laziest senior developer. YAGNI-first. Stdlib-first. No unrequested abstractions. The shortest diff that works.
Packaged as 7 pure SKILL.md skills for Hermes Agent by tensakulabs/hermes-ponytail.
1. Prerequisites¶
- Hermes Agent installed and running
- Hermes skills CLI (
hermes skills install)
No Python dependencies, no tools, no API keys - these are pure skill files.
2. Installation¶
Install all 7 skills at once:
hermes skills install tensakulabs/hermes-ponytail/ponytail --yes
hermes skills install tensakulabs/hermes-ponytail/ponytail-review --yes
hermes skills install tensakulabs/hermes-ponytail/ponytail-audit --yes
hermes skills install tensakulabs/hermes-ponytail/ponytail-debt --yes
hermes skills install tensakulabs/hermes-ponytail/ponytail-gain --yes
hermes skills install tensakulabs/hermes-ponytail/ponytail-help --yes
Or install individual skills as needed.
Skills land in ~/.hermes/skills/ and are auto-discovered on the next restart.
3. Skill Reference¶
ponytail - Lazy Senior Dev Mode¶
Core mindset skill. Activates YAGNI-first thinking: - No unrequested abstractions (no factories, no patterns without a problem) - Stdlib before dependencies - Shortest diff that works - Delete code before adding code - Question every dependency, every abstraction layer
Load via: skill_view(name='ponytail')
ponytail-review - Diff Review¶
Review any diff for over-engineering. Output: one line per thing to delete, one line per thing to simplify.
Usage: Load before reviewing PRs or when asked to review code.
ponytail-audit - Whole-Repo Audit¶
Systematic audit of the entire repo for over-engineering. Produces a ranked list of what to cut - most impactful simplifications first.
Usage: Load when doing architectural health checks or before major refactors.
ponytail-debt - Tech Debt Ledger¶
Harvests every ponytail: shortcut comment in the codebase into one tracked debt ledger. Comments like # ponytail: this abstraction is premature become tracked items.
Usage: Run periodically to maintain visibility of intentional shortcuts and planned simplifications.
ponytail-gain - Impact Scoreboard¶
Measures Ponytail's impact: less code, lower costs, faster builds. Trackable metrics over time.
Usage: Load for monthly retrospectives or to demonstrate simplification ROI.
ponytail-help - Quick Reference¶
One-page reference of all Ponytail modes and skills with one-liner descriptions.
4. CorpusIQ Use Cases¶
- Agent skill development - Use
ponytail-reviewwhen reviewing new skill SKILL.md files before publishing - Infrastructure code -
ponytail-auditon workstation and automation scripts - Growth ops - Keep automation scripts minimal;
ponytailmindset prevents over-engineered workflows - Docs management - Audit the corpusiq-docs repo structure for unnecessary complexity
5. Limitations¶
- On-demand only - Hermes skills load when triggered, unlike Claude Code/Codex hooks that run every turn. If you want Ponytail governing every Hermes turn, append the ruleset from
ponytail/SKILL.mdinto Hermes' base system prompt or build it as a Hermes plugin (pre_llm_callhook). - No code execution - These are pure behavioral skills. They change how the agent thinks, not what tools it has.
- Not for urgent hotfixes - Ponytail's simplification-first mindset can slow down emergency patches where speed trumps elegance. Disable or unload for critical incidents.
6. Verification¶
After installation, verify the skills are available:
hermes skills list | grep ponytail
Expected output: all 6 ponytail skills listed (ponytail, ponytail-review, ponytail-audit, ponytail-debt, ponytail-gain, ponytail-help).
7. Uninstall¶
hermes skills uninstall ponytail
hermes skills uninstall ponytail-review
hermes skills uninstall ponytail-audit
hermes skills uninstall ponytail-debt
hermes skills uninstall ponytail-gain
hermes skills uninstall ponytail-help
← Skills Catalog | Hermes Ponytail Repo →
Related¶
- Skills Marketplace - 946+ community skills with setup guides
- Hermes Changelog - Latest agent releases and features
Integration with CorpusIQ¶
CorpusIQ is the connector layer that lets AI assistants read the systems a business already runs on - accounting, payments, CRM, email, files and analytics - and answer with the source named. It is read-only by design: it retrieves data and never writes to, or acts in, a connected system, which is what makes it safe to point at live accounts.
This page is part of the CorpusIQ directory, the reference we maintain for the MCP server and agent-skill ecosystem. Run alongside CorpusIQ, a third-party server like this one can be combined with a company's own business data in a single answer.