hermes-field-kit

Hermes Field Kit

Field-tested, open-source skills for Hermes Agent.

Tags: hermes-agent · hermes-ai · ai-agents · agent-skills · developer-tools · automation · open-source · tonysimons_

Hermes Field Kit is a curated collection of reusable workflows that have earned their place through real, repeated use. It is intentionally not a bulk skill dump. Experimental entries are explicitly marked while they accumulate the field evidence required for stable promotion.

Current status

Version 1.0.1 is the current tagged release. The 1.1.0 release branch expands the working catalog to thirteen stable skills plus six experimental skills.

The catalog includes private-by-default analytics, source-locked writing, evidence discipline, session continuity, Hermes-native work decomposition, post-build review, and an operational Field Kit organized around:

inspect -> diagnose -> plan -> execute -> verify -> hand off

The repository applies the same admission rule to every future stable skill.

The admission rule

A skill belongs in the stable catalog only when all three statements are true:

1. It solves a real task.
2. It has been used in an actual workflow.
3. Another person can reproduce the intended behavior from the repository.

Generated filler, speculative prompts, thin wrappers, and untested skill bundles do not qualify. Experimental skills must have a concrete use case, explicit limitations, behavior tests, and a defined path to stable promotion.

What every published skill must provide

• A Hermes-compatible SKILL.md
• Direct tap discovery under skills/<skill-name>/
• Clear triggers and counter-triggers
• Sanitized examples from realistic workflows
• Behavior-oriented test cases
• Known limitations and failure modes
• Platform and tool requirements
• Independent versioning
• No embedded credentials, private data, or personal configuration

See the skill specification for the complete contract.

Repository map

skills/<skill-name>/       Published tap-discoverable skills
catalog.json               Machine-readable index and category metadata
templates/skill-template/  Nonfunctional authoring scaffold
schemas/                   Machine-readable catalog and test schemas
scripts/validate.py         Dependency-free repository validator
scripts/validate_release_wave.py  Published-skill and public-tree hardening validator
tests/                     Validator contract tests
docs/                      Installation, design, testing, and release policy
.github/                   Contribution forms, dependency updates, and CI

Published skills

• what-have-we-done-today experimental: Scan today's sessions, kanban boards, and cron runs across all profile stores and write an append-friendly daily markdown recap.
• dont-lie-to-me experimental: Separate observation, sourced claims, user reports, inference, unknowns, and contradictions before strong factual or completion claims.
• hermes-change-review experimental: Review implementation independently for intent fidelity, repository quality, and actual verification evidence.
• hermes-environment-migration: Safely migrate Hermes environments with staged archives, integrity manifests, secret separation, selective imports, verification, and rollback.
• hermes-gateway-doctor: Diagnose gateway failures from process, adapter, credential-posture, log, delivery, and persistence evidence without automatic repair.
• hermes-kanbanize experimental: Turn settled conversation/spec context into a Hermes-native Kanban graph with acceptance criteria, blocking edges, a verified frontier, and execution gating.
• hermes-profile-audit: Compare a profile's declared responsibilities with its actual tools, skills, persistence, access, and observed behavior without rewriting it.
• hermes-session-handoff experimental: Build a portable continuation packet with verified state, artifacts, blockers, re-verification needs, and an exact fresh-session prompt.
• hermes-skill-audit: Audit global and profile-local skills for dependencies, frontmatter, usage integrity, cron references, duplicates, and upstream drift.
• hermes-skill-consolidate experimental: Safely consolidate or restructure overlapping skills with read-only planning, scope-bound approval, rollback snapshots, staged writes, and post-change verification.
• hermes-stack-doctor: Discover the installation architecture, delegate to focused evidence contracts, and report a GREEN, YELLOW, or RED stack verdict without repairs.
• hermes-token-audit: Audit token usage and cost with live schema discovery, aggregate-first privacy, and clear separation between estimates and provider billing.
• hermes-update-doctor: Investigate update failures by separating remote drift, repository divergence, process locks, stale caches, partial installs, and runtime mismatches.
• interview-me: Ask one high-value question at a time, inspect available sources before questioning, and stop when more questions would not change the next action.
• omarchy-theme-maker: Derive a contrast-checked Omarchy theme palette from any image, install it as the theme background, and apply it with omarchy theme set.
• oss-tool-trust-audit: Read source and release machinery, treat popularity as context rather than proof, and separate technical legitimacy from adoption fit.
• pre-build-feature-audit: Run a read-only duplicate check across source, history, branches, issues, pull requests, roadmaps, and contributor guidance.
• repo-readiness-audit: Determine whether a Git repository is ready for development, release, handoff, or contribution using independent evidence from repository and collaboration surfaces.
• x-analytics-import: Validate, normalize, import, and compare X Analytics CSV exports through a repeatable private-by-default workflow.
• x-post-writer: Draft, rewrite, and repurpose short-form X content with source fidelity, format routing, and claim verification.

Design principles

• Field-tested over fashionable.
• Trust over volume.
• Process predictability over ornamental prose.
• Precise routing pointers over vague skill descriptions.
• Progressive disclosure over giant always-loaded instructions.
• Single sources of truth over copied environment facts.
• Reproducible behavior over promises of guaranteed outcomes.
• Safe defaults over environment-specific shortcuts.
• Tap compatibility over aesthetically tidy but undiscoverable nesting.

Read Design Principles for the reasoning behind these rules.

Installing skills

Install any published skill by its repository-qualified identifier:

hermes skills inspect asimons81/hermes-field-kit/hermes-stack-doctor
hermes skills install asimons81/hermes-field-kit/hermes-stack-doctor --yes

Replace hermes-stack-doctor with any name from the published-skills list. Start a new Hermes session after installation because discovery may be cached for the lifetime of an existing session.

See Installation for inspection, installation, update behavior, removal, manual installation, and platform paths.

Validation

python scripts/validate.py
python -m unittest discover -s tests -v
python scripts/validate_release_wave.py

The repository validator checks tap layout, catalog agreement, frontmatter, behavior cases, supporting paths, JSON validity, and common secret patterns. The hardening validator runs contract tests for every published skill, runs bundle validators where supplied, parses every Python file, checks relative Markdown links, rejects generated artifacts, and scans the public tree for private or credential-bearing material.

Contributing

Start with CONTRIBUTING.md. New skill proposals use the dedicated GitHub issue form and must include evidence of real use without exposing private information.

Versioning and releases

The repository and catalog use Semantic Versioning. Individual skills carry their own SemVer version in SKILL.md. No release tag will be created solely for the scaffold.

See Release Process and Compatibility.

Security

Never submit API keys, session tokens, private analytics exports, customer data, unpublished credentials, or personal configuration. Report suspected exposure privately as described in SECURITY.md.

Relationship to Hermes Agent

This is an independent community repository for skills compatible with Hermes Agent. It is not an official Nous Research repository and does not imply endorsement.

License

Apache License 2.0. See LICENSE.