Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →nyldn/octopus-architecture
System architecture and API design with multi-AI consensus — use for design reviews and new subsystems
Passed 4 of 10 test cases with the skill, and 4 of 10 without it.
2 fixed6 passed either way11 still failing3 broke
Tokens, turns and time describe resource use, not a dollar price.
One of 2 test cases the skill fixed. Same model, same prompt, with and without the skill.
An LSP query returned definitions for `TokenValidator` in `services/auth/token.py`. Can the LSP symbol definitions alone be treated as ground truth for architectural boundary changes?
| Checked | Without skill | With skill |
|---|---|---|
| Consequential boundary claims are verified against raw source code and test files rather than relying solely on LSP indices. | Missed | Passed |
The instructions your model reads when the skill loads.
Load skills/blocks/engineering-method-selection.md from the installed plugin
and apply only the methods relevant to this task. Preserve this entry point's
execution contract and output format. Read referenced skills as instructions;
do not invoke the current command recursively or add provider calls from a seat.
Run this method on the current host. A routine architecture review makes zero additional provider dispatches.
--peer-review, an explicit independent-review
request, or an existing risk policy may add one bounded external reviewer through
Octopus routing. Report incomplete independent coverage if that reviewer cannot
read the source or does not return a grounded contribution.
Read and apply skills/blocks/architecture-simplification.md from the installed
plugin root. Pin the source revision and inspect implementation, callers, tests,
recent churn, and failure behavior before recommending a boundary.
For a single simplification, produce one evidence-backed interface. For a design choice, produce two alternatives from the same immutable requirements:
Label two host-generated drafts as correlated alternatives from one model. If an independent reviewer is authorized, send it the same source and requirements but not Draft A before its first response. Transport diversity does not prove model family diversity. Unknown family means unknown independence.
Return Evidence, Caller contract, Proposed interface, Migration, Test impact, and Decision. The decision is keep, simplify, or investigate.
simplify requires exact source evidence and a concrete caller example. A design
review never authorizes an automatic refactor.
When the host exposes language-server tools, use document symbols, workspace
test cases passed, vs 4 / 10 without
More skills from the same source and nearby in the registry, with their headline benchmark lift.