engflow-poc
Skill0 starsChanged 17 days agoMITView on GitHub
Test a technical premise through a minimal, bounded experiment. Exercise the real mechanism and return reproducible evidence with explicit limits. Use on explicit experiment request. Other Engflow entrypoints reuse the shared experiment method within their own execution authority. Return results to the caller or conversation; save a durable report only when requested. Not production implementation.
In this repository13 other files it gives its agents.
- AGENTS.mdAGENTS.md
- diagram-aestheticsSkill
- engflow-clarifySkill
- engflow-code-reviewSkill
- engflow-designSkill
- engflow-handoffSkill
- engflow-implementSkill
- engflow-planSkill
- engflow-refactor-assessSkill
- engflow-researchSkill
- engflow-specifySkill
- engflow-tasksSkill
- semver-bumpSkill
The file
We split this skill into the rules it gives an agent
1 rule. About 0 are common, 0 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈0
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 31
When explicitly asked to retain the experiment, reuse its existing record or the specified/project location.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 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.
Files
Engflow PoC
Resolve a consequential uncertainty with the smallest adequate experiment, not a partial implementation of the whole feature. No other Skill or document is required.
Engflow is installed as a complete group. An explicit PoC request starts this entrypoint; follow-up trials reuse its granted scope. Read and apply the shared experiment method before any probe. Other entrypoints use that reference directly, not an implicit PoC invocation. A request to inspect, plan, design or save results alone authorizes no execution.
Results and persistence
Return results to the coordinating caller, or directly in conversation when used independently. The caller integrates consequential evidence into its own authorized sources; PoC does not edit the caller's Design or progress record. Calls do not automatically create research/experiment reports. Experiment execution may still need authorized temporary files; distinguish those from requested durable storage.
When explicitly asked to retain the experiment, reuse its existing record or the
specified/project location. Otherwise default to docs/poc/NNN-short-name/, with
report.md and, when needed, runnable material under code/. Keep only the code,
inputs and results needed to reproduce and interpret the evidence; no empty code
directory or fixed report template is required. Report scope, source/environment,
actual commands/results, meaningful failed trials, conclusions and limitations.
For a new experiment record, choose the next number above allocated PoC records,
starting at 001 with at least three digits. Keep identities stable, do not fill
gaps, and recheck collisions before writing. Research and feature numbering remain
independent. Continue the same experiment in place, retaining relevant history;
preview does not allocate a record. Saving supplied results does not rerun the
experiment or turn historical evidence into a fresh observation. Link related
sources without creating a feature or changing its navigation. Return retained paths
and disclose missing reproducibility material rather than inventing it.
Execution and coordination
The shared experiment method governs evidence quality, execution authority, stopping and live-state reconciliation. It grants no production edits or publication. For an explicitly coordinated PoC task, also preserve the following ownership rules.
Under a coordinator, obey its scope and exclusive-write rules. Workers start or resume read-only until a fresh grant; they do not delegate further or edit the coordinator's design/progress record. Return evidence for the coordinator's decisions.
Quiesce experiment processes before returning where possible; identify outstanding processes and whether they can still write. A timeout or result message does not prove they stopped. Reconcile live state before retrying.
Report retained artifacts and side effects; clean up only experiment-owned resources within authority, preserving needed evidence. Do not implement production changes, commit, push, publish or manage branches/worktrees.
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
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.

