docs: September 2026 Copilot coding standards

.github/copilot-instructions.md

@@ -22,3 +22,7 @@ Read [`AGENTS.md`](../AGENTS.md) and [`docs/DESIGN-PIVOT.md`](../docs/DESIGN-PIV
 ## Key paths
 
 `policy.cedar`, `policy.cedarschema`, `protect-mcp.config.json`, `packages/cedar-daemon`, `apps/secure-gateway`, `packages/crypto-utils`, `apps/admin-dashboard`.
+
+## Coding standards
+
+Follow path-specific files in [`.github/instructions/`](instructions/). Copilot code review loads [`.github/skills/code-review/`](skills/code-review/).

.github/instructions/cedar.instructions.md

@@ -0,0 +1,10 @@
+---
+applyTo: "**/*.cedar,**/*.cedarschema"
+---
+
+# Cedar coding standards (September 2026)
+
+- Keep authorization fail-closed. A missing policy, schema mismatch, or unavailable PDP is a deny.
+- Do not introduce a silent fallback from enforce to shadow.
+- Validate against the checked-in schema (`policy.cedarschema` or the service equivalent).
+- Pair policy edits with the existing Cedar validation script or tests.

.github/instructions/tests.instructions.md

@@ -0,0 +1,15 @@
+---
+applyTo: "**/*.{test,spec}.ts,**/*.{test,spec}.tsx,**/tests/**/*.ts,**/tests/**/*.tsx"
+---
+
+# Test standards (September 2026)
+
+- Ship behavior with a test in the existing layout (co-located `*.test.ts` or `tests/`).
+- Do not skip, delete, or weaken verify, lint, typecheck, or adversarial gates to land a change.
+- Do not invent a passing gate from prose. Run the documented verify command.
+- Fixtures are synthetic. Never commit real PII, PHI, payroll, or credentials.
+
+## This repository
+
+- Definition of Done: `./scripts/harness/verify.sh`.
+- Cedar: `bash scripts/cedar-validate.sh`.

.github/instructions/typescript.instructions.md

@@ -0,0 +1,21 @@
+---
+applyTo: "**/*.ts,**/*.tsx"
+---
+
+# TypeScript coding standards (September 2026)
+
+- Write TypeScript, not new application JavaScript. Leave existing tooling `.js` / `.mjs` / `.cjs` files alone.
+- Place imports at the top of the module. Do not use inline `import()` in function bodies except for a documented circular-dependency or optional-runtime case.
+- Prefer early returns over nested conditionals.
+- On `switch` over a discriminated union or enum, handle every variant. Use a `never` check in `default` so newly added variants fail at compile time.
+- Do not introduce `any` in new production code. Prefer `unknown` plus narrowing. Do not use non-null assertions to silence `strict` or `noUncheckedIndexedAccess`.
+- Match this repository's existing formatter and linter. Do not add a second style system.
+- Keep public unions narrow. Do not widen a literal union to `string` without a spec change.
+- Do not weaken fail-closed, human-in-the-loop, tenancy, or verify gates to make types compile.
+
+## This repository
+
+- Authorization and Ed25519 receipts stay on the MCP tool-call path.
+- Do not convert authorize, kill-switch, PDP, principal-signature, or production-profile paths to fail-open.
+- Shadow mode may log-without-block only when `protect-mcp.config.json` `mode` is explicitly `shadow`.
+- Do not treat this tree as a prompt-injection classifier.

.github/skills/code-review/SKILL.md

@@ -0,0 +1,28 @@
+---
+name: code-review
+description: "Review FidusGate PRs for fail-closed Cedar authorize and Ed25519 receipt integrity. Use on pull requests that touch policy.cedar, the secure gateway, crypto-utils, or the admin console. Flag silent enforce-to-shadow fallbacks."
+---
+
+# Copilot code review — FidusGate
+
+Use this skill when reviewing a pull request in this repository.
+
+FidusGate issues **signed Ed25519 receipts for MCP tool calls**.
+
+- Reject fail-open authorize, kill-switch, PDP, or production-profile paths.
+- Reject silent fallback from enforce to shadow.
+- Do not treat Cedar work as a prompt-injection classifier.
+- Verify with `./scripts/harness/verify.sh` and `bash scripts/cedar-validate.sh` when policies change.
+
+
+## Always flag
+
+- Secrets, `.env` values, private keys, or real personal data in the diff
+- Weakened or skipped verify / lint / typecheck / adversarial gates
+- Invented success (prose claiming a gate passed with no command output)
+- Fail-open authorization, skipped human approval, or agents recording `--actor user`
+
+## Never request
+
+- Drive-by major upgrades, formatter churn, or unrelated refactors
+- Softening honesty disclaimers or certification claims

AGENTS.md

@@ -41,3 +41,9 @@ Do not convert authorize, kill-switch, PDP, principal-signature, production-prof
 ## Honesty
 
 Keep demo and mock surfaces labeled as demo (local keys, JSON datastore, simulated syscalls/OIDC). Do not add sidecar features to compete with OpenFirma, Vectimus, Symbiont, or Permit Cedar Agent. Deepen receipts and the console.
+
+## Coding standards (September 2026)
+
+Path-specific Copilot instructions: [`.github/instructions/`](.github/instructions/).
+Repository-wide Copilot file: [`.github/copilot-instructions.md`](.github/copilot-instructions.md).
+Copilot code review skill: [`.github/skills/code-review/SKILL.md`](.github/skills/code-review/SKILL.md).

llms.txt

@@ -24,3 +24,6 @@ FidusGate does not compete as a generic zero-trust agent governance platform. Op
 - [MCP 2026-07-28 migration](https://github.com/SafetyMP/FidusGate/blob/main/docs/mcp-2026-07-28-migration.md)
 - [OWASP MCP Top 10 ADR](https://github.com/SafetyMP/FidusGate/blob/main/docs/adr/0001-owasp-mcp-top-10.md)
 - [GitHub storefront playbook](https://github.com/SafetyMP/FidusGate/blob/main/docs/community/github-presentation.md)
+- [.github/copilot-instructions.md](.github/copilot-instructions.md): repository-wide Copilot coding standards
+- [.github/instructions/](.github/instructions/): path-specific Copilot coding standards (September 2026)
+- [.github/skills/code-review/SKILL.md](.github/skills/code-review/SKILL.md): Copilot code review skill