Skip to content

opensiro/vsm-harness-skills / skills/build-harness-cognee-first

build-harness-cognee-first

Verified

Build a new agent harness through verifiable Cognee-first architectural discovery, source inspection, engineering decisions, and traceable implementation; runtime benchmarking is optional.

FileKindSize
SKILL.mdskill md14.4 KB
VERSIONother6 B
references/case-study.mdmarkdown2.8 KB
references/decision-trace.mdmarkdown7.4 KB
references/run-contract.mdmarkdown7.2 KB

references/case-study.md

View on GitHub
# Flash API case: a reproducible development-process example, not an architecture recipe

Historical source: https://github.com/opensiro/harness-build-example-flash-api/tree/5202f0a9330b54da69bce657cbea8e206ffe5658

This case study is external to the present Skills repository. Its code and original receipts are **unchanged** and permanently referenced at the historical merge SHA. Its intended result is the **construction decision trace**; scripted task evaluation was extra evidence, not a required benchmark or the skill's success criterion.

Observed chronology in PR #1:
1. A Python dependency repair task and source selection were fixed before architectural discovery. Index commit 14a2182e2d6d0ecd851cef118906431b2c851aaa; connector d176484e85da55c84450d67040cb4a2d096353c7, with a disclosed one-line sharded-path fix.
2. A deterministic 128/577 included-assessment sample was imported into a fresh local Cognee 1.6.4 dataset: 2,661 graph nodes, 1,765 edges, 768 exact grades; independent audit PASS.
3. Seven preregistered CHUNKS_LEXICAL queries returned native hits (including irrelevant negative controls). **Only the six VSM function sections of each full ingested assessment were lexical-search chunks**, not the entire Markdown; TLDR.md was not used.
4. A post-query shortlist and nine upstream code inspections at pinned SHA preceded the architecture decision. The resulting small adaptation was a Carbon-inspired observed post-write test gate.
5. Local Python tests passed 9/9 at final review. A scripted task-specific double resulted in baseline final tests FAIL and adapted final tests PASS. **Real Fast/Flash model run NOT RUN; no model-quality uplift established.**

Primary evidence:
- [Discovery and its coverage](https://github.com/opensiro/harness-build-example-flash-api/blob/5202f0a9330b54da69bce657cbea8e206ffe5658/docs/cognee-discovery.md)
- [Raw Cognee results](https://github.com/opensiro/harness-build-example-flash-api/blob/5202f0a9330b54da69bce657cbea8e206ffe5658/receipts/discovery/native-lexical-results.json)
- [Independent audit](https://github.com/opensiro/harness-build-example-flash-api/blob/5202f0a9330b54da69bce657cbea8e206ffe5658/receipts/audit/independent-audit.json)
- [Candidate comparison](https://github.com/opensiro/harness-build-example-flash-api/blob/5202f0a9330b54da69bce657cbea8e206ffe5658/reports/candidate-comparison.md)
- [Decision Trace](https://github.com/opensiro/harness-build-example-flash-api/blob/5202f0a9330b54da69bce657cbea8e206ffe5658/traces/decision-trace.jsonl)
- [Final evaluation](https://github.com/opensiro/harness-build-example-flash-api/blob/5202f0a9330b54da69bce657cbea8e206ffe5658/reports/evaluation.md)

For a new game/research/coding task, run *fresh* retrieval and comparisons. Carbon or test gating is not a mandatory or privileged choice.
reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis.Stars: 472kUpdated Claude CodeCodexVerifiedInstall
diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness.Stars: 322kUpdated Claude CodeCodexVerifiedInstall
dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependenciesStars: 322kUpdated Claude CodeCodexVerifiedInstall
executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is availableStars: 322kUpdated Claude CodeCodexVerifiedInstall