agentleFS
Cody-SimsCody-Sims/agent-skills/skills/shadow-dream/SKILL.md

shadow-dream

Skill0 starsChanged 4 months agoMITView on GitHub

Proposes evidence-linked future-state architecture options from Shadow observations and accepted constraints. Use when asked to dream, compare architecture options, define a target state, or propose migration and rollback; not for product specifications, implementation tasks, accepted decision recording, or implementation.

The file

We split this skill into the rules it gives an agent

3 rules. About 0 are common, 3 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈3

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 0

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Load relevant context: Read the declared Shadow entry point/schema, observations, accepted decisions, relations, anchors, and evidence.

    line 34
  2. Distinctive:

    Compare/recommend: Use identical drivers/constraints.

    line 57
  3. Distinctive:

    Return/persist: Use repository-required fields first, retaining all information above.

    line 63

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Shadow Dream

Produce a reviewable architecture proposal, not an approved decision. Inputs: question/horizon, repository instructions and declared .shadow/ README/index/schemas/lifecycle/approval conventions, evidence-linked current observations, accepted constraints, and optional goals/risk/deadline/migration limits. Repository declarations govern graph representation; missing inputs remain unknown.

Hard Boundary

Remain proposal-only: never edit code, configuration, infrastructure, tests, or implementation plans; mark accepted; change accepted decisions/history; or claim human approval. Refuse acceptance/implementation requests and offer only the proposal, approval gate, or later handoff. Persist at most one proposed artifact only on explicit request, using the declared path/schema. If this requires other artifact or accepted-record changes, return in chat and report the conflict.

Workflow

  1. Bound. State title, scope, horizon, outcome, non-goals, and status proposed.
  2. Load relevant context. Read the declared Shadow entry point/schema, observations, accepted decisions, relations, anchors, and evidence. Preserve IDs and lifecycle meanings; do not survey unrelated records.
  3. Classify evidence. Separate Verified inputs supported by observations, constraints, instructions, schemas, code/tests/docs, issues/PRs/commits; Inferences labeled and linked to supporting inputs; and Unknowns covering absent, contradictory, stale, or unverified facts. Cite workspace- relative paths or stable references. Missing rationale is always unknown. Inference cannot become an accepted constraint; proposals cannot prove their own acceptance.
  4. Extract drivers. List evaluation criteria and mark constraints as accepted, requested, inferred, or unknown. Only accepted architecture claims are mandatory on that basis; current hard user scope, security, and budget limits remain binding without graph acceptance. Keep optional preferences distinct and surface authority conflicts rather than silently overriding them.
  5. Generate options. Produce at least two materially distinct architectures, not implementation variants. Include status quo/incremental change when a credible baseline.
  6. Describe every option. Include target structure/responsibilities, supporting evidence and satisfied constraints, assumptions/unknowns, benefits/tradeoffs, operational/organizational consequences, risks/failure modes and evidence to reduce them, migration stages/gates/observable exit criteria, and rollback boundaries/triggers/method/irreversible steps.
  7. Compare/recommend. Use identical drivers/constraints; distinguish verified comparison facts from judgment, with no unsupported scores or false precision. Recommend an option, hybrid, or no change; explain evidence fit, remaining tradeoffs, and unknowns that could reverse it.
  8. Review gate. List open questions, needed evidence, required reviewers, and the repository-declared human approval artifact. Keep proposed.
  9. Return/persist. Use repository-required fields first, retaining all information above. Default to the complete proposal in chat. Only if explicitly requested and the writing boundary holds, validate one proposed artifact against the declared schema and report its path without promotion.
  10. Hand off only with evidence of approval. Locate and cite the declared human approval artifact, not model language. Give shadow-architecture the proposal, approval reference, evidence links, unknowns, and requested recording action; never perform that action here. Without the artifact, stop at the proposal.

Response

Report status, chat/persistence, sources, unknowns, option count, recommendation, migration/rollback, open questions, approval requirement, and explicit refusal of requested acceptance/history rewriting/implementation. Use requirements-and-spec-writing for product requirements, use cases, acceptance criteria, or interface specifications; planning-and-task-breakdown for implementation tasks after approval.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.