agentsclimarketplace

Prompt engineering

Skill codegraphtheory/hermes-profile-template/skills/prompt-engineering

Prompt-to-repo authoring system with one-sentence generation, demo assets, diagrams, and installable Hermes Agent profile distributions.

Install
npx -y skills add codegraphtheory/hermes-profile-template --skill prompt-engineering

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

What its author says it does

Copied from the file, not written here

Turn a simple profile idea into a mature, production-quality Hermes profile prompt and generation brief.

The file declares its own license as MIT. That is the author’s claim about this one file, and it is not the same thing as the license GitHub reports for the repository, which is listed with the other numbers below.

SKILL.md

3.5 KB, as published. Nobody here has run it

Prompt Engineering for Hermes Profiles

Use this skill when a user gives a short idea such as "make me a database migration reviewer" and wants a high-quality Hermes profile distribution.

The goal is not to make a longer prompt for its own sake. The goal is to turn a simple sentence into a mature profile specification that can generate an installable, safe, useful Hermes profile repo.

Inputs

Minimum input:

  • A one-sentence profile idea.

Useful optional input:

  • Target user or team.
  • Domain and workflow.
  • Expected artifacts.
  • Tools or integrations.
  • Risk level and data sensitivity.
  • Desired tone and output format.
  • Repository output path.

Workflow

1. Extract intent

From the user's simple sentence, infer:

  • job to be done
  • target user
  • trigger situations
  • expected outputs
  • likely tools
  • likely risks
  • what must be refused

Ask at most three clarifying questions only if missing information would materially change the generated profile. Otherwise proceed with explicit assumptions.

2. Write a mature profile prompt

Create a Mature Profile Prompt with these sections:

  1. Profile name and slug.
  2. Mission.
  3. Target users.
  4. Primary workflows.
  5. Trigger patterns.
  6. Inputs the profile expects.
  7. Outputs the profile must produce.
  8. Tool-use policy.
  9. Safety boundaries and refusals.
  10. Required skills.
  11. Environment variables, if any.
  12. Verification and smoke-test expectations.
  13. Repository output requirements.

Keep it specific enough that another agent could generate the profile repo without more context.

3. Convert mature prompt to params YAML

Map the mature prompt into templates/profile.params.yaml fields:

  • name
  • display_name
  • description
  • toolsets
  • env_requires
  • principles
  • scope
  • refusals
  • output_contract
  • github_topics

Add a profile_prompt field containing the mature prompt so the generated repo can preserve the design rationale in docs/profile-prompt.md when supported by the generator.

4. Generate the repository

Run:

python3 scripts/generate_profile.py --params <params.yaml> --output <target-dir>

Then add any profile-specific bundled skill that materially improves the profile. Keep skills narrow, procedural, and domain-specific.

5. Verify installability

Run:

python3 <target-dir>/scripts/validate_profile.py <target-dir>

When Hermes is available, also run:

hermes profile install <target-dir> --name <smoke-name> --yes --force

Quality checklist

  • The mature prompt is concrete, not generic.
  • The generated profile has one clear job.
  • The generated profile includes safety boundaries.
  • The profile does not claim fake integrations, credentials, audits, or community links.
  • Required env vars are documented in .env.EXAMPLE only.
  • At least one domain-specific bundled skill is added when the workflow benefits from reusable procedure.
  • Validation passes.
  • Smoke install passes when Hermes is available.

Output format

When finishing, report:

  • Simple sentence received.
  • Mature profile prompt path, if written.
  • Params YAML path.
  • Generated repository path.
  • Validation command and exact result.
  • Smoke-install command and exact result, or why skipped.
  • Next publish command.

Keep looking

Skills are one crate of 328,083. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.