Agentic engineering: lint/CI/supply-chain policy + deckard-core hardening - #8
Merged
Conversation
One-time normalization so the new `cargo fmt --all --check` CI gate passes. Formatting only — no logic changes (rustfmt never alters semantics).
Move the lint policy into the manifest and enforce it in CI so agents see the same rules at `cargo check` time that CI gates: - [workspace.lints] + [lints] in both crates; clippy.toml (msrv 1.95, disallowed-methods: mem::forget/thread_rng); checked-in rustfmt.toml. - deckard-core: #![forbid(unsafe_code)] + #![deny(unwrap_used, expect_used, get_unwrap, indexing_slicing, panic, mem_forget)]. The untrusted-byte parser is refactored to bounds-checked .get()/.first()/try_into(); the two startup-fatal expects carry a documented #[allow]; #[must_use] on the secret types (Vault/UnlockedVault). - deny.toml + non-blocking cargo-deny CI job (advisories/licenses/bans/sources). - CI: add fmt --check, default-feature clippy, cargo test, and --locked. - docs/AGENTIC-ENGINEERING.md (shareable upstream with Deck) + CLAUDE.md/AGENTS.md. Verified green: clippy (default + tray), cargo test 13/13, cargo fmt --check. Cross-model reviewed by codex GPT-5.5. Tier 2/3 backlog: #7.
This was referenced Jun 7, 2026
Vault::open collapses every unlock failure (parse, KDF, wrong-pass, AEAD, IO) to one generic message so the unlock screen can't reveal whether a vault is tampered/corrupt vs the passphrase wrong. Also adds `just core` (fast GPUI-free engine loop) and frozen v1 decode-compat fixtures.
Integrate main's virtual-workspace restructure (deckard-app/core/contract/signerd), verified reads (Helios), the key-less shield path, and the signer daemon WITH the agentic-engineering hardening (workspace lints, deckard-core restriction lints + forbid-unsafe, keystore no-oracle Vault::open contract, CI gates, docs). Conflicts resolved: Cargo.toml (virtual workspace + workspace.lints), ci.yml + justfile (combined coverage + --locked/fmt-check/cargo-deny/just core), lib.rs/eth.rs (main + re-applied lint block & the two startup expect allows), shell.rs/main.rs/welcome.rs (main; the unlock oracle is now handled by the daemon's UnlockOutcome::BadPassphrase collapse, so the keystore Vault::open contract is complementary). deny.toml + AGENTS.md + docs/AGENTIC-ENGINEERING.md adapted to the new 4-crate workspace. Verified green: clippy --workspace -D warnings (default + tray), cargo test --workspace (81 passed, 1 ignored), cargo fmt --check. Merge reviewed cross-model by codex GPT-5.5.
hellno
pushed a commit
that referenced
this pull request
Jun 14, 2026
Second decision round (2026-06-14): - Ambition: harvest aligned attributes, do not climb the Stage ladder - Hardware-wallet signing (#4): out of scope (deliberate non-goal) - Browser integration (#25): deferred; MCP agent is the sole connection surface; app isolation (#11) reframed as per-agent/session isolation - Account abstraction (#26): EIP-7702 research spike (on-chain agent policy), batching (#27) contingent on it - Multi-identity (#8/#9): a post-v0 goal with non-correlation UX Rewrites the Stages section as reference-only.
hellno
pushed a commit
that referenced
this pull request
Jun 14, 2026
Validate the code-fit layer from a four-cluster read of the actual codebase: per-item crate placement, seam-vs-expansion, size, and risk, plus an EIP-7702 spike brief. Key findings: - deckard-core already has the right seams (multi-account derivation, reserved enclave flag, alloy/7702-ready builder-agnostic broadcast) - Un-gating Send is UI-only (daemon path tested) and the dependency hub - deckard-contract (frozen) is the higher-ceremony expansion point; batch RevokeApproval/Intent::Batch/session-token changes together - Re-tier #13 RPC-before-first-request UP to L (provider spawns with DEFAULT_RPC before the auth gate); re-tier #8/#9 multi-identity DOWN (core already derives accounts) Reworks the sequencing tiers with the measured sizes.
hellno
pushed a commit
that referenced
this pull request
Jun 14, 2026
Second decision round (2026-06-14): - Ambition: harvest aligned attributes, do not climb the Stage ladder - Hardware-wallet signing (#4): out of scope (deliberate non-goal) - Browser integration (#25): deferred; MCP agent is the sole connection surface; app isolation (#11) reframed as per-agent/session isolation - Account abstraction (#26): EIP-7702 research spike (on-chain agent policy), batching (#27) contingent on it - Multi-identity (#8/#9): a post-v0 goal with non-correlation UX Rewrites the Stages section as reference-only.
hellno
pushed a commit
that referenced
this pull request
Jun 14, 2026
Validate the code-fit layer from a four-cluster read of the actual codebase: per-item crate placement, seam-vs-expansion, size, and risk, plus an EIP-7702 spike brief. Key findings: - deckard-core already has the right seams (multi-account derivation, reserved enclave flag, alloy/7702-ready builder-agnostic broadcast) - Un-gating Send is UI-only (daemon path tested) and the dependency hub - deckard-contract (frozen) is the higher-ceremony expansion point; batch RevokeApproval/Intent::Batch/session-token changes together - Re-tier #13 RPC-before-first-request UP to L (provider spawns with DEFAULT_RPC before the auth gate); re-tier #8/#9 multi-identity DOWN (core already derives accounts) Reworks the sequencing tiers with the measured sizes.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Adds an agentic-engineering guardrail layer so coding agents (and humans) write reliable Rust fast: the lint policy moves into the manifest (
[workspace.lints]+clippy.toml+ a checked-inrustfmt.toml) and CI now enforcesfmt --check,clippy -D warningson both feature configs,cargo test,--locked, and a non-blockingcargo-denysupply-chain gate. The security enginedeckard-coreis locked down with#![forbid(unsafe_code)]plus crate-level denies ofunwrap/expect/panic/raw-indexing in non-test code, its untrusted-byte parser refactored to bounds-checked accessors, and#[must_use]on the secret types. A one-timecargo fmtbaseline (separate commit) normalizes pre-existing formatting drift. Rationale and the deliberately-rejected rules live indocs/AGENTIC-ENGINEERING.md(written to share upstream with the Deck starter), withCLAUDE.md/AGENTS.mdupdated and the Tier 2/3 backlog tracked in #7. All changes verified green (clippy both configs, 13 tests, fmt) and cross-model reviewed by codex GPT-5.5.