docs: resolve all rustdoc warnings (clippy and compiler already clean)

crates/tinyagents-harness-tools/src/scratchpad/store.rs

@@ -1,5 +1,5 @@
 //! Persistence for the per-thread scratchpad, on the harness
-//! [`Store`](tinyagents_harness::store::Store).
+//! [`Store`].
 //!
 //! One serialized [`Scratchpad`] per thread under [`SCRATCHPAD_NAMESPACE`],
 //! keyed by the hex-encoded thread id, written `load → apply → put` under a

crates/tinyagents-harness-tools/src/state_block/mod.rs

@@ -16,7 +16,7 @@
 //!   and those of its last finished task ([`crate::scratchpad`]), written
 //!   through the `todo` tool's note arguments;
 //! - **the ledger**: facts derived from the run's own tool calls with no model
-//!   involved ([`ledger`]): the last few failed calls with their exit code and
+//!   involved (`ledger`): the last few failed calls with their exit code and
 //!   first stderr line, executables a shell reported missing, and files
 //!   written by write/edit tools. Facts correct themselves: a later success of
 //!   the same command clears its failure, and a later successful command that

crates/tinyagents-harness/src/agent_loop/tools.rs

@@ -2417,7 +2417,7 @@ impl<State: Send + Sync, Ctx: Send + Sync> AgentHarness<State, Ctx> {
     /// Reconciles unresolved tool-effect-ledger rows (B5) before resuming a
     /// run from durable transcript `messages`.
     ///
-    /// A crash between [`Self::record_tool_effect_started`] and the matching
+    /// A crash between `Self::record_tool_effect_started` and the matching
     /// settle write (or between the settle write and the tool result being
     /// folded into `messages`) leaves a `started` row a resumed run must
     /// resolve one way or another before it can safely feed `messages` back
@@ -2456,7 +2456,7 @@ impl<State: Send + Sync, Ctx: Send + Sync> AgentHarness<State, Ctx> {
     /// mistake: any call id in it is skipped even if its ledger row is
     /// (unexpectedly) still `started` — e.g. the `Deferred` settle write
     /// above failed and was only logged (settle writes are best-effort, see
-    /// [`Self::record_tool_effect_settled`]). [`AgentHarness::resume_deferred`]
+    /// `Self::record_tool_effect_settled`). [`AgentHarness::resume_deferred`]
     /// passes the ids `results` is about to answer; any other caller — a host
     /// reconciling a genuine crash, where no `results` exists at all — passes
     /// an empty set.

crates/tinyagents-harness/src/context/mod.rs

@@ -444,7 +444,7 @@ impl<Ctx> RunContext<Ctx> {
     /// child-specific keys.
     ///
     /// This keeps the child's `Ctx` type identical to the parent's, which is
-    /// what makes propagating [`Self::host_authority`] sound: the type-erased
+    /// what makes propagating `Self::host_authority` sound: the type-erased
     /// authority installed by a hosted invocation is keyed to the exact
     /// `(State, Ctx)` pair it was constructed for, and this method is the only
     /// place that carries it forward. A recursive call that needs a

crates/tinyagents-harness/src/cost/mod.rs

@@ -101,7 +101,7 @@ fn select_tier(pricing: &ModelPricing, input_tokens: u64) -> Option<&PriceTier>
 /// non-cached/non-reasoning remainder only.
 ///
 /// When `pricing` declares [`ModelPricing::tiers`], the tier matching this
-/// call's `usage.input_tokens` (see [`select_tier`]) supplies the input,
+/// call's `usage.input_tokens` (see `select_tier`) supplies the input,
 /// output, cache-read, and cache-write rates, falling back to the flat
 /// [`ModelPricing`] fields for any rate the tier leaves unset. Reasoning
 /// tokens are always priced at the flat

crates/tinyagents-harness/src/middleware/library/research_budget.rs

@@ -39,7 +39,7 @@ pub const WEB_BUDGET_EXHAUSTED_INSTRUCTION: &str = "The direct web research budg
 /// Consecutive failed web calls after which web access looks blocked.
 const BLOCKED_AFTER_FAILURES: usize = 2;
 
-/// The one-time note sent after [`BLOCKED_AFTER_FAILURES`] failed web calls in a
+/// The one-time note sent after `BLOCKED_AFTER_FAILURES` failed web calls in a
 /// row (a blocked network, an offline sandbox).
 pub const WEB_BLOCKED_NOTE: &str = "Web access looks blocked: the last web calls failed. Do not keep retrying web tools; continue the task with your other tools and the information already available.";
 

crates/tinyagents-harness/src/middleware/library/tool_policy.rs

@@ -21,7 +21,7 @@ impl ToolAllowlistMiddleware {
 
     /// Returns `true` if `name` is on the allowlist.
     ///
-    /// Delegates to [`crate::tool::toolset::tool_name_allowed`] — the exact
+    /// Delegates to `crate::tool::toolset::tool_name_allowed` — the exact
     /// membership test [`crate::tool::toolset::FilteredToolSet::allowing`]
     /// uses — so this middleware and its `ToolSet` counterpart cannot drift
     /// (`docs/runtime-comparison/pydantic-ai.md` §4).

crates/tinyagents-harness/src/middleware/library/unmet_deliverable.rs

@@ -86,7 +86,7 @@ fn has_extension(segment: &str) -> bool {
 }
 
 /// Absolute paths with a file extension that `text` names, in order, deduped
-/// and capped at [`MAX_CANDIDATES`].
+/// and capped at `MAX_CANDIDATES`.
 ///
 /// Absolute only: a relative token in prose ("see config/settings.yml") is far
 /// more often a reference than a deliverable, and a relative path cannot be

crates/tinyagents-harness/src/no_progress/post_compaction.rs

@@ -40,7 +40,7 @@ impl PostCompactionGuard {
     }
 
     /// Records one successful call; `repeating` says whether the pair had
-    /// already recurred [`REPEATING_AT`] times. Before a compaction only
+    /// already recurred `REPEATING_AT` times. Before a compaction only
     /// repeating pairs are remembered. Returns `true` when the guard is armed
     /// and this call repeats a remembered pair (and disarms, so one compaction
     /// flags at most one repeat).

crates/tinyagents-harness/src/retry/failover.rs

@@ -121,7 +121,7 @@ impl FailoverReason {
     }
 }
 
-/// The failover table. Pure; see the [module docs](self) for the rationale.
+/// The failover table. Pure; see the module docs for the rationale.
 pub fn decide(reason: FailoverReason, state: FailoverState) -> FailoverDecision {
     use FailoverReason::*;
     match reason {

crates/tinyagents-harness/src/steering/mod.rs

@@ -431,7 +431,7 @@ impl SteeringHandle {
 /// - When `ctx` has no [`SteeringHandle`], returns
 ///   [`SteeringOutcome::Continue`] without emitting anything.
 /// - Only commands addressed to this run are drained: see
-///   [`SteeringHandle::drain`] and [`SteeringHandle::for_child`]. Commands
+///   [`SteeringHandle::drain`] and `SteeringHandle::for_child`. Commands
 ///   addressed to a different run stay queued for it.
 /// - Each command is checked **individually** against the run's
 ///   [`SteeringPolicy`]. A disallowed command is rejected on its own — an

crates/tinyagents-harness/src/steering/types.rs

@@ -24,7 +24,7 @@ use tinyinference_llm::message::Message;
 /// Every [`SteeringHandle`] clone shares one underlying queue (so an
 /// orchestrator can hand a single handle to a deeply nested tree and still
 /// reach any run in it), but each level of the tree only *drains* the entries
-/// addressed to itself — see [`SteeringHandle::for_child`]. Without this, a
+/// addressed to itself — see `SteeringHandle::for_child`. Without this, a
 /// command meant for the orchestrating run could be consumed by whichever
 /// sub-agent happened to reach a checkpoint first.
 #[derive(Clone, Debug, PartialEq, Eq, Serialize, Deserialize)]
@@ -302,7 +302,7 @@ pub struct PauseState {
 /// A plain `clone()` is a bare alias: it shares this handle's identity
 /// (`run_id`/`is_root`) as well as its queue, so it drains exactly the same
 /// commands this handle would. When a run spawns a child,
-/// [`crate::context::RunContext::child`] calls [`SteeringHandle::for_child`]
+/// [`crate::context::RunContext::child`] calls `SteeringHandle::for_child`
 /// (not `clone`) so the child only drains commands addressed to it or to
 /// [`SteeringTarget::All`] — see that method's docs.
 #[derive(Clone)]

crates/tinyagents-harness/src/summarization/task_state/ledger.rs

@@ -29,7 +29,7 @@ const MAX_COMMAND_CHARS: usize = 200;
 impl TaskLedger {
     /// Folds the facts in `messages` into this ledger (which may carry facts
     /// from earlier checkpoints). Files keep most-recent order; commands keep
-    /// the most recent [`MAX_COMMANDS`].
+    /// the most recent `MAX_COMMANDS`.
     pub fn absorb(&mut self, messages: &[Message]) {
         if self.original_task.is_none() {
             self.original_task = messages

crates/tinyagents-harness/src/title/mod.rs

@@ -115,7 +115,7 @@ pub fn collapse_whitespace(input: &str) -> String {
 /// Rules applied (in order):
 /// - split on anything that is not alphanumeric (punctuation, quotes, markdown,
 ///   and whitespace all become word breaks)
-/// - drop [`FILLER_WORDS`], unless that would leave nothing
+/// - drop `FILLER_WORDS`, unless that would leave nothing
 /// - keep the first [`THREAD_TITLE_MAX_WORDS`] words, and stop early rather
 ///   than exceed [`THREAD_TITLE_MAX_CHARS`]
 ///

crates/tinyagents-harness/src/tool/deferred/mod.rs

@@ -1,5 +1,5 @@
 //! Deferred tool calls: the typed, resumable "the loop paused on a tool"
-//! output (A2). See [`types`] for the vocabulary and the agent-loop docs for
+//! output (A2). See `types` for the vocabulary and the agent-loop docs for
 //! how the loop produces and consumes it.
 
 mod types;

crates/tinyagents-harness/src/tool/toolset/external/types.rs

@@ -8,7 +8,7 @@ use tinytools::ToolSpec;
 /// Mirrors Pydantic AI's `defer_loading`/deferred-tools model
 /// (`docs/runtime-comparison/pydantic-ai.md` §3.4, `pi.md` §4.1's
 /// transcript-carried tool-loadout changes). It advertises
-/// [`Self::schemas`] to the model like any other toolset, but never runs one
+/// `Self::schemas` to the model like any other toolset, but never runs one
 /// of them locally.
 ///
 /// # Host-integration contract

crates/tinyagents-harness/src/tool/toolset/mod.rs

@@ -173,7 +173,7 @@ where
 
 /// Bridges one [`ToolSet`]-owned tool into the [`crate::tool::ToolDispatch`]
 /// the agent loop's admission path already speaks, so a name only a toolset
-/// chain exposes (not [`crate::tool::ToolRegistry::model_dispatch`]) can be
+/// chain exposes (not `crate::tool::ToolRegistry::model_dispatch`) can be
 /// admitted and executed through the exact same call path as a directly
 /// registered tool — timeout policy, injected-argument handling, schema
 /// validation, and every other admission step in `agent_loop/tools.rs` apply

crates/tinyagents-harness/src/tool/toolset/prepared/mod.rs

@@ -27,7 +27,7 @@ impl<State: Send + Sync, Ctx: Send + Sync> PreparedToolSet<State, Ctx> {
 
     /// Convenience constructor: keeps only the schemas a per-schema
     /// predicate accepts, ignoring `ctx`. Built on
-    /// [`retain_matching_schemas`], the same retain logic
+    /// `retain_matching_schemas`, the same retain logic
     /// [`crate::middleware::library::DynamicToolSelectionMiddleware`] uses.
     pub fn filtering(
         inner: Arc<dyn ToolSet<State, Ctx>>,

crates/tinyagents-harness/src/tool/toolset/prepared/types.rs

@@ -22,7 +22,7 @@ pub type SchemaTransform<Ctx> =
     Arc<dyn Fn(&RunContext<Ctx>, Vec<ToolSchema>) -> Vec<ToolSchema> + Send + Sync>;
 
 /// [`ToolSet`] adaptor that applies a caller-supplied, per-step
-/// [`SchemaTransform`] to an inner toolset's declared schemas.
+/// `SchemaTransform` to an inner toolset's declared schemas.
 ///
 /// This is the same seam [`crate::tool::SchemaPreparation`] occupies for
 /// provider projection, generalised to a caller-supplied closure that is

crates/tinyagents-orchestration/src/run_progress_ledger/mod.rs

@@ -1,6 +1,6 @@
 //! Best-effort durable observation of one agent turn's progress stream.
 //!
-//! The ledger writes each [`AgentProgress`](tinyagents_harness::progress::AgentProgress)
+//! The ledger writes each [`AgentProgress`]
 //! as an append-only run event and keeps the run status aligned with turn
 //! completion. It also preserves subagent task-to-parent-call correlation for
 //! host projections whose later lifecycle events omit the parent call id.

crates/tinyagents-registry/src/pricing/mod.rs

@@ -35,7 +35,7 @@ fn normalize(model: &str) -> String {
 /// Match order: exact canonical id → id with a leading `vendor/` segment
 /// stripped → longest canonical id that the request extends with a snapshot
 /// decoration only (`gpt-5.4-mini-2026-05-01`, `claude-sonnet-4-5-20250929`,
-/// `gemini-2.5-pro-preview-06-05`; see [`is_decorated_id`]). Returns `None`
+/// `gemini-2.5-pro-preview-06-05`; see `is_decorated_id`). Returns `None`
 /// for unknown models — callers should fall back to a tier estimate.
 ///
 /// Any other suffix names a different model, not a snapshot of a listed one: