hermes-runtime
Use when any pstack skill spawns work, picks a model, asks a question, or reads a transcript.
Skill by klgraham
Install
Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.
npx agentmarket install skill/klgraham-hermes-runtime
Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.
Get the weekly diff
The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.
About
Use cases: agent-building
Reviews
Write a reviewNo reviews yet. Installed it? Reviews come from verified installers only. How reviews work
Permissions
What this skill can do and what it asks for, from its SKILL.md. Checked Oct 25, 2026.
- Tools
- No tools found
- Secrets
- None asked for
- Install scripts
- None found
Network hosts it talks to 1
codeload.github.com· found in install command
Files it touches
~/.claude/skills/hermes-runtime
What's inside
- Skill name
hermes-runtime- Description
- Use when any pstack skill spawns work, picks a model, asks a question, or reads a transcript. Hermes binding for this plugin. poteto-agent is a skill, not a profile.
- License
- MIT
- Instructions
- 92 lines in SKILL.md · view on GitHub
Files (1)
- SKILL.md
SKILL.md (first lines)
# Hermes runtime This plugin is the Hermes port of poteto's pstack. It is not a Cursor plugin. Do not create a Hermes profile named poteto, poteto-agent, or pstack. When a skill still mentions Task, subagent_type, AskQuestion, ~/.cursor, a cloud agent, or cursor-team-kit, obey this file instead of that sentence. ## Invocation - Entry point: `/pstack:poteto-mode` or `skill_view(name='pstack:poteto-mode')`. - Sibling skill: `skill_view(name='pstack:<name>')`. - Playbook: `skill_view(name='pstack:poteto-mode', file_path='playbooks/<name>.md')`. - Principle leaf: `skill_view(name='pstack:principle-<kebab>')`. - Hermes has no sticky custom mode. Once the user invokes poteto-mode, keep applying it until they opt out. The chat that received poteto-mode owns the playbook, the todo list, and every spawn. Do not spawn a child to run the playbook. ## Model config Read `$HERMES_HOME/pstack/models.md` before the first spawn. Resolve `HERMES_HOME` from the environment, otherwise the parent of `hermes config path`, otherwise `~/.hermes`. A missing file, or a missing role line, means `inherit-parent`. `inherit-parent` and `auto` mean the parent chat model. Hermes slugs do not encode reasoning. Do not append `-xhigh`, `-fast`, `-max`, or similar. The `# reasoning:` line is the effort for this run (`none`, `low`, `medium`, `high`, `xhigh`). Do not override a reasoning level the user already set. `delegate_task` cannot pin a model per child. Every child inherits the parent model, or the global `delegation.model` pin if one is set. Do not edit `config.yaml` to fake a per-role model. When a role's slug differs from the parent model, and setup-pstack confirmed that slug, run that seat as its own process: ``` hermes chat -q '<self-contained prompt>' -m <slug> --provider <provider> ``` Put the full brief in the prompt. Background it and wait for stdout. If the slug was not confirmed, run the seat as `inherit-parent` via `delegate_task` and say so. A panel whose every entry is `inherit-parent` is one model. Say that model diversity was unavailable. Do not duplicate the parent model to fake a two-model panel. Same-model N-way generation is still valid when the skill asks for it. ## Delegation Code-writing and helper delegates inside a playbook: 1. `delegate_task`, one `tasks` entry per delegate, parallel entries in one call. 2. Child context must say: load `pstack:poteto-agent` before any work, then load `pstack:poteto-mode` and read Principles. You are a delegate. You do not own the playbook. Absolute path to poteto-mode `SKILL.md`: pass it. 3. Children cannot call `clarify`. If they need the user, they return the question and the parent decides. 4. Children do not see this conversation. Pass paths, SHAs, and the brief. Do not inline large files. 5. A child summary is a self-report. Review the diff yourself. 6. Fresh child per new round, unless the work needs that child's checkout, uncommitted changes, or a process it still runs. Steer with `delegate_task` action `steer`. Stop with action `stop`. 7. Top-level `delegate_task` runs in the background and the result returns between turns. After dispatch, finish unrelated work, give a one-line status, and end the turn. Do not poll transcripts. Routed skills (how, why, arena, swarm, interrogate, reflect, architect) set their own worker shape. Do not force poteto-agent onto a readonly reviewer. Still pass a self-contained brief. Read-only means the brief forbids edits. Hermes does not have a readonly subagent flag. ## Questions `AskQuestion` does not exist. Use `clarify` for a genuine product or preference call. Put options in `choices`, not in the question text. Under a full-autonomy grant, do not ask. Apply a default, act, and report it. ## Todos Open playbook steps with the `todo_list` tool before task-specific items. A skipped step stays, marked skipped, with a one-line reason. ## Isolation No Cursor cloud agents. No `environment: cloud`. No `cloud_base_branch`. Parallel writers get their own `git worktree`. Read-only reviewers do not need one. ## Tools that are not here - `deslop` / cursor-team-kit: use `pstack:unslop` on the prose and the diff, then `pstack:no-comments`. - `control-cli`: drive the CLI with the terminal. Reproduce before handing work to the user. - `control-ui`: drive the UI with `browser_exec` or `computer_use`. Reproduce before handing work to the user. - Cursor `create-skill`: write the `SKILL.md` yourself. Frontmatter needs `name` and `description`. - Bugbot: ordinary review comments (GitHub review threads, CI, any bot that commented). Triage each. The rubric is `skill_view(name='pstack:poteto-mode', file_path='references/bugbot-triage.md')`. - Origin CLI: not available. Land with `gh`. - `/loop`: a Hermes cron job, or keep going in this session. There is no Cursor loop command. ## Evidence and transcripts - Source control: `git` and `gh`. - Other evidence: tools already in this session, plus `hermes mcp list`. Do not look for a `mcps/` directory. - Session history: `session_search`. Session files live under `$HERMES_HOME/sessions/`. Do not read `~/.cursor/projects` or `.cursor/skills`. - A skill counts as used when this session loaded it with `skill_view`, or the transcript shows that load. Not when a Cursor `Read` of `~/.cursor/plugins` appears. ## Verification Prove the change against the real artifact. A child saying it passed is not proof. ## Not ported The Benny automation pack is Cursor automations plus Slack. It is not in this plugin. Do not copy it into `.cursor/automations`.
From klgraham/pstack-hermes, MIT.
More from klgraham/pstack-hermes
- 0.565score
comment-sicko
klgraham
Use when no-comments spawns a comment review.
SkillAFreshness: AClaude Code +9 - 0.559score
poteto-agent
klgraham
Use when a poteto-mode delegate starts work.
SkillAFreshness: AClaude Code +9 - 0.541score
how
klgraham
Use for "how does X work", code walkthroughs before changing something, and placement / ownership / layering questions ("where should this live", "which…
SkillAFreshness: AClaude Code +9 - 0.541score
create-verification-skill
klgraham
Generate a project-local verification skill that drives your app the way a user does — any language, framework, or platform.
SkillAFreshness: AClaude Code +9 - 0.541score
architect
klgraham
Sketch types, signatures, and module structure before code, then stay in the loop while implementation fills in.
SkillAFreshness: AClaude Code +9 - 0.541score
interrogate
klgraham
Use for "interrogate", "adversarial review", "multi-model review", "challenge this", "stress test this code", "find blind spots", or "tear this apart".
SkillAFreshness: AClaude Code +9 - 0.539score
correct
klgraham
Find the mistakes agents keep repeating in this repo and make each one impossible.
SkillAFreshness: AClaude Code +9 - 0.539score
benchmark-checklist
klgraham
Vet a perf measurement (limiter, tuning, limits, errors, repeatability, relevance, and whether the work happened) before you report or act on it.
SkillAFreshness: AClaude Code +9 - 0.536score
automate-me
klgraham
Use for "automate me", "create/update/refresh my -mode skill", "turn/capture my preferences or working style into a skill", or wanting agents to follow how…
SkillAFreshness: AClaude Code +9 - 0.536score
blast-radius
klgraham
Find what a change could break somewhere else before it ships, beyond the diff, and prove the one fact it's safe because of by running real code instead of…
SkillAFreshness: AClaude Code +9 - 0.536score
maintain-verification-skill
klgraham
Periodic pass that keeps a project's verification skill and feature map honest: parallel source readers per feature, one live session driving every feature…
SkillAFreshness: AClaude Code +9 - 0.536score
arena
klgraham
Spawn N parallel candidates at the same task, pick a base, graft the strongest parts of the losers into it.
SkillAFreshness: AClaude Code +9
Signals
- Stars
- 0
- Forks
- 0
- Open issues
- 0
- Last commit (this folder)
- Oct 15, 2026 (13 days ago)
From GitHub (fetched Oct 24, 2026).
Details
- License
- MIT
- Repository
- github.com/klgraham/pstack-hermes
- First seen
- Oct 24, 2026
- Last checked
- Oct 24, 2026
- Source
- GitHubgithub-graphql
More from klgraham
- 0.565score
comment-sicko
klgraham
Use when no-comments spawns a comment review.
SkillAFreshness: AClaude Code +9 - 0.559score
poteto-agent
klgraham
Use when a poteto-mode delegate starts work.
SkillAFreshness: AClaude Code +9 - 0.541score
create-verification-skill
klgraham
Generate a project-local verification skill that drives your app the way a user does — any language, framework, or platform.
SkillAFreshness: AClaude Code +9 - 0.541score
how
klgraham
Use for "how does X work", code walkthroughs before changing something, and placement / ownership / layering questions ("where should this live", "which…
SkillAFreshness: AClaude Code +9 - 0.541score
architect
klgraham
Sketch types, signatures, and module structure before code, then stay in the loop while implementation fills in.
SkillAFreshness: AClaude Code +9 - 0.541score
interrogate
klgraham
Use for "interrogate", "adversarial review", "multi-model review", "challenge this", "stress test this code", "find blind spots", or "tear this apart".
SkillAFreshness: AClaude Code +9
Cite this page
hermes-runtime skill. Agent Mart. https://agentmart.dev/skills/klgraham-hermes-runtime (last updated 2026-10-24)