FDE Consultants Protocoles turns Claude, Codex, Cursor, Hermes and more into a disciplined operator — it scopes the real problem, builds it, and certifies every result against verifiable, file-level evidence.
Pilots stall, "looks done" isn't done, and no one can verify the result after the demo. FDE Consultants Protocoles encodes Palantir-grade field-engineering discipline so your agent produces work it can defend — with a score you can audit.
Every engagement runs the loop. Nothing is "done" until its evidence trail resolves on disk. See the protocols →
Recon the actual codebase, then a 6-question decomposition — no vague "use AI".
Surface ≥3 failure modes before shipping. Doubt the method before, not after.
Every claim ties to a real file, line, or test. Evidence is verified, not asserted.
An FDE Assurance Score, written to a public, hash-chained registry anyone can verify.
The brain. Any agent installs it and ships certified work.
7 local tools — recon, decompose, roi, scientific-search, evals, ontology, trust-score.
Free curriculum overview — the operator method, in the open.
Public, hash-chained log of certifications. Verify anything.
# clone, then one command — every detected runtime git clone https://github.com/selectess/fde-consultants-protocoles cd fde-consultants-protocoles && bash install.sh ✅ Claude Code → ~/.claude/skills/ ✅ Codex → ~/.agents/skills/ ✅ Hermes → ~/.hermes/skills/ ✅ openclaw → ~/.openclaw/workspace/skills/
Open-source. Local-first. Verifiable. Installs in under a minute.