agentsclimarketplace

Run

Skill hyuniiiv/quantum-agent/skills/run

Install
npx -y skills add hyuniiiv/quantum-agent --skill run

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

One thing to look at

  • 1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

QuantumAgent — Execute a declarative quantum experiment defined in a YAML file. Parses spin/singularity/spawn config from experiment.qa.yaml and orchestrates the full workflow. The CI/CD-friendly entry point to QuantumAgent — quantum experiments as code.

SKILL.md

9.1 KB, as published. Nobody here has run it

cosmos:run

Execute a declarative quantum experiment from a YAML file. This is the programmable layer of QuantumAgent — experiments become version-controlled artifacts that can be reviewed, diffed, and re-run.

While /cosmos spawn is imperative ("do this now"), /cosmos run is declarative ("here's the experiment, execute it"). The same goal can be captured as code, committed to git, and re-executed identically.

Trigger

/cosmos run <path-to-yaml>

Example:

/cosmos run experiments/rate-limiting.qa.yaml
/cosmos run my-experiments/auth-review.qa.yaml

YAML schema (v1)

# experiment.qa.yaml — schema version 1

experiment: <kebab-case identifier>     # required, string
version: 1                              # optional, schema version, default 1

# Optional macro-layer setup (applied BEFORE spawn)
spin:                                   # optional block
  name: <string>                        # required if spin block present
  type: <string>                        # optional
  description: <string>                 # optional
  constraints:                          # optional list
    - <constraint string>
    - <constraint string>

singularities:                          # optional block, list
  - name: <string>                      # required
    invalidates:                        # required list
      - <pattern>
    trigger: <string>                   # optional
    description: <string>               # optional

# Required: the spawn configuration
spawn:                                  # required block
  goal: <string>                        # required

  # Two forms — pick ONE. Cannot use both.
  #
  # Form A — simple list (uniform model):
  strategies:                           # 2-5 unique entries
    - <strategy>
    - <strategy>

  # Form B — verbose (per-cosmos model, v3.3+):
  cosmos:                               # 2-5 unique entries
    - { strategy: <strategy>, model: <model-name> }
    - { strategy: <strategy>, model: <model-name> }

  entanglement: <mode>                  # optional: none|passive|active|strict
                                        # default: passive
  models:                               # optional list, only with strategies form
    - <model-name>                      # length must match strategies
    - <model-name>

Validation rules:

  • Top-level keys: experiment, version, spin, singularities, spawn. Reject unknown keys.
  • experiment must be present and a non-empty string.
  • spawn must be present with goal (string) and EITHER strategies (list of 2-5 unique strings) OR cosmos (list of 2-5 unique entries with strategy field) — not both.
  • spawn.entanglement if present must be one of none, passive, active, strict.
  • spawn.models (optional, only with strategies form) must match strategies length.
  • spawn.cosmos entries each require strategy; model is optional per entry.
  • spin.name required if spin block exists.
  • Each singularities[] entry needs name and invalidates.
  • Strategy uniqueness (across forms) check enforces Pauli Exclusion at the declaration layer.

Execution Steps

Step 1 — Detect repo root

git rev-parse --show-toplevel

Store as <repo_root>.

Step 2 — Resolve YAML path

If the argument is a relative path, resolve against <repo_root>. If the file does not exist:

❌ Experiment file not found: <path>
   Usage: /cosmos run <path-to-yaml>
   Templates available at: experiments/_template.qa.yaml

Then stop.

Step 3 — Parse and validate YAML

Read the file. Parse as YAML. Apply validation rules from the schema above. On error:

❌ Invalid experiment YAML at <path>:
   <specific validation error>

   See schema: experiments/_template.qa.yaml or /cosmos run --schema

Then stop. Common errors to surface explicitly:

  • Missing experiment → "Required field experiment missing at top level."
  • Missing spawn → "Required block spawn missing."
  • Missing spawn.goal → "Required field spawn.goal missing."
  • Empty/short spawn.strategies → "spawn.strategies must list 2 to 5 unique strategies."
  • Duplicate strategies → "Pauli Exclusion: strategy <name> listed more than once."
  • Unknown entanglement value → "spawn.entanglement must be one of: none, passive, active, strict."
  • Unknown top-level key → "Unknown key <key> at top level. Allowed: experiment, version, spin, singularities, spawn."

Step 4 — Report parsed experiment

Before executing anything, summarize what will happen:

🧪 Experiment: <experiment-name> (schema v<version>)

   Spin operations:        <yes if spin block / no>
   Singularity operations: <N events / none>
   Spawn:
     Goal:           <goal>
     Strategies:     <strategy1>, <strategy2>, <strategy3>
     Entanglement:   <mode>

   Executing...

Step 5 — Apply spin (if spin block present)

If the YAML contains a spin: block, invoke the same logic as /cosmos spin:

  • Read existing <repo_root>/.quantum/project/spin.json if any
  • Merge the YAML's spin block with existing values (the YAML wins on conflicts; established is preserved)
  • Write <repo_root>/.quantum/project/spin.json
  • Output: ✨ Spin applied: <name>

If no spin: block in the YAML, skip silently.

Step 6 — Apply singularities (if singularities block present)

For each entry in the singularities: list:

  • Generate ISO 8601 timestamp <ts> (current time, or YAML-provided ts if present)
  • Build the JSON event:
    {"name":"<name>","ts":"<ts>","trigger":"<trigger>","invalidates":[...],"description":"<description>"}
    
  • Append to <repo_root>/.quantum/singularities/events.jsonl
  • Output: ☄️ Singularity declared: <name>

If no singularities: block, skip silently.

Step 7 — Execute spawn

Build the equivalent /cosmos spawn arguments from the spawn: block:

  • --goal "<goal>"
  • --strategies "<comma-joined strategies>"
  • --entanglement <mode> (if not default)
  • --models "<comma-joined>" (if either Form A models: or Form B cosmos[].model is present)

For Form B (cosmos verbose form): extract strategy and model from each entry into parallel lists for the underlying spawn call.

Then invoke the same execution flow as /cosmos spawn (Steps 2 through 8 of skills/spawn/SKILL.md).

The macro context loaded in Step 2.5 of spawn will now include the spin and singularities applied in Steps 5 and 6 above.

Step 8 — Echo experiment provenance

Add a run marker insight to each cosmos's namespace AFTER spawn dispatches but BEFORE agents begin:

echo '{"type":"run","experiment":"<experiment-name>","schema_version":<version>,"yaml_path":"<resolved-path>","ts":"<now>"}' \
  >> <repo_root>/.quantum/<name>/insights.jsonl

This makes every insight traceable back to the originating experiment file. Auditors can answer "which experiment produced this insight?" by reading the first line of the JSONL.

Step 9 — Auto-observe completion

Identical to /cosmos spawn Step 8 — when all agents return, /cosmos observe runs automatically.

Why declarative experiments

Reproducibility — the same YAML always produces the same spawn configuration. Discoveries depend on agent runs, but the experiment design itself is fixed and version-controlled.

Review — experiment files can be reviewed in pull requests. "Why are we testing these three strategies?" becomes a discussion in git, not a Slack thread.

Re-execution/cosmos run experiments/auth-review.qa.yaml always means the same thing. Useful for periodic audits ("run the security audit every quarter").

CI/CD — quantum experiments can run on a schedule, in response to triggers, or as part of release gates.

Composition — multiple experiments can share spin and singularities through the same YAML file or evolve them across files.

When to use /cosmos run vs /cosmos spawn

Use /cosmos spawn whenUse /cosmos run when
Quick one-off explorationThe experiment will be repeated
Strategies are still being chosenStrategies are stable
You're prototyping the questionThe question is well-defined
No need for audit trail beyond .quantum/Provenance ("which experiment produced this?") matters

Both ultimately call the same underlying spawn logic — /cosmos run is /cosmos spawn with a declarative façade and macro-layer integration.

Schema versioning

The version: field declares the YAML schema version. Current: 1. If absent, 1 is assumed.

Future schema versions may add fields. Older schema files remain readable — /cosmos run reads version to determine which validation rules to apply.

Format reference

See experiments/_template.qa.yaml for an annotated template, and experiments/rate-limiting.example.qa.yaml for a real-world example.

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.