[JustinKovacich]

Writing pass over crates/uds_services to the "Writing for the reader" standard (CONTRIBUTING.md, #84, which this PR is stacked on). Prose only: rustdoc, comments, the Cargo.toml comments and the README. No code line changes and no test renamed; identifier.rs and encode.rs had nothing to change.

Before / after (prose lint): hard-fail phrases 8 / 0 (one "load-bearing", five "deliberately", two ticket references); history 5 / 0; em-dash 448 / 255 (every one in a touched sentence); comment blocks of 12+ lines 103 / 70. The 70 that remain are the uds_server! and uds_client! macro docs with their doctests and compile_fail blocks, the crate overview, the Encode/Sink re-export example, the UdsTransport obligation list, the pipeline's Figure 5 stage list and Table I.2 bullets, and the per-method contracts (# Arguments, # Errors, # Cancel safety) of the thirty Client calls, Server::new/step/run, and every service-trait handler; the narration inside them was cut, the sections stay.

Fixed while reading: the README said "the transport binding is an optional feature", which contradicted the crate doc and Cargo.toml (no features; the application names the binding). The crate doc and README pointed readers at test files that do not ship in the crate's rendered docs; those references are gone.

Removed, and the reasoning it held (this PR is its record; nothing is filed elsewhere):

• Why a server has three buffers (the handler borrows the in-flight buffer, so the driver waits on the concurrent one to see the tP2_Server deadline) and a client two (a repeat re-sends the encoded request while the caller holds the response). Candidate Rationale: for UDSSVC_ARCH_0017/0041.
• Why EMPTY is an associated const (trait methods cannot be const), the ConstStaticCell versus StaticCell::init versus plain static choices, and the 8,824-byte frame measurement cross-referenced from uds_on_ip.
• The tracking issue for a terminal Err and step's cancel safety (uds_stack#19), and the "milestone-1 limit" framing on six items; the behaviour statements stay.
• Why peers = 1 only (one slot for a selecting response awaiting confirmation), kept to one sentence. Candidate Rationale: for UDSSVC_ARCH_0040.
• reply_address is physically addressed by observation rather than by a clause of ISO 14229-2 9.6 (UDSS_LLR_0026); the driver as "the component no standard describes, which went a design cycle with no owner"; the Deferred::merge ordering argument in full.
• Clause 8.7.5's pseudo-code agreeing with the stage order (UDSSVC_ARCH_0006); why the pipeline takes no context struct (UDSSVC_ARCH_0035); why assembly is a macro and not a blanket impl, the $ty:ty follow-set note, and the rustfmt 1.9.0 non-convergence detail (UDSSVC_ARCH_0013).
• The arguments for the transport seam's shape: a subslice rather than a length, a DataTooLong variant rather than a flag, why Debug is the one error bound, why there is no inbound_max, why Closed is an edge. The contracts stay.
• Why select2 is private; the seal's "lock, not a wall" account and the derivation argument; why service traits have no defaults and MAY_RESPOND_PENDING none; the per-service motivations behind each MAY_RESPOND_PENDING (a flash write, a routine, an erase).
• Session, security, transfer, DTC and routine trait detail that explained rather than specified (the four-state machine's "corrected topology", SecurityLevel's pairing arithmetic, Figure 30's three-methods argument, "open question 5", "decision 4").
• Client docs: the storage-mirroring argument, why Records cannot fail step by step, the Option-versus-bool reasoning, the quoted 15.2.3.2 sentence, and two driver references to architecture open questions.
• tests/composition.rs: the DTC_RECORD_LEN it replaced and the retyping of the handler seam; Cargo.toml: the alloc/std features once forwarded to uds_protocol.

Issue URL

No issue; the pass follows the standard agreed in #84.

Testing

• python3 tools/prose_lint.py crates/uds_services --fail --show 50 exits 0 with every hard-fail row at 0; CI's Comment prose job now runs --strict crates/uds_services.
• cargo fmt --all --check; cargo clippy -p uds_services --all-targets --all-features -- -D warnings; cargo nextest run -p uds_services --all-features (402 passed) and the doctests (13 + 3); cargo doc -p uds_services --all-features --no-deps with -D rustdoc::broken_intra_doc_links -D warnings.
• git diff -U0 ... | grep -vE '^[-+]\s*(//|#|\*|/\*|$)' prints nothing: no code line changed.

🤖 Generated with Claude Code