feat(flow): harden codex workflow routing and review ops

Codex workflows in Flow had grown useful but still had weak edges around crash recovery, review handoff, and prompt inspection. This change makes the wrapper path more deterministic and cheaper to reopen: Codex sessions can now be listed quickly from the state DB after a restart, optional thread metadata is read schema-safely, and explicit workflow routes for review/commit/sync/session references stay compact instead of relying on raw chat inference.

It also turns PR feedback and skill work into inspectable runtime surfaces. Flow now emits richer review artifacts, exposes resolve and skill sync/reload APIs for the daemon and myflow, logs daily Codex/daemon activity, and dedupes sync output more robustly. The intent is better robustness first: recover state cleanly after crashes, avoid duplicated or stale context, and make the exact routing/context decisions visible before Codex starts.

Author: nikivdev

docs/commands/ai.md

@@ -15,6 +15,8 @@ f ai
 # Provider-specific list/read
 f ai claude list
 f ai codex list
+f ai codex sessions
+f ai codex sessions --path ~/repos/mark3labs/kit
 f ai cursor list
 f codex resolve "latest"
 f codex resolve "https://linear.app/.../project/.../overview" --json
@@ -97,6 +99,30 @@ Behavior:
 This keeps prompt cost flat unless Flow has a strong reason to recover or unroll context.
 Use `f codex doctor` to confirm whether wrapper transport, runtime skills, and context budgets are actually active for the current repo.
 
+### Codex sessions after a crash or restart
+
+If your Mac restarts and you lose the live Codex terminals, use:
+
+```bash
+f ai codex sessions
+f ai codex sessions --path ~/repos/mark3labs/kit
+```
+
+Behavior:
+
+- lists recent Codex sessions for the current path
+- sorts by the last message timestamp descending
+- shows the stable session id plus a preview of the latest message
+- the numeric index matches `continue`, so you can reopen quickly
+
+Examples:
+
+```bash
+f ai codex continue 1 --path ~/repos/mark3labs/kit
+f ai codex continue 019cd046 --path ~/repos/mark3labs/kit
+f ai codex sessions --path ~/repos/mark3labs/kit --json
+```
+
 ### Optional `flow.toml` resolver config
 
 You can teach `f codex open` and `f codex resolve` to unroll repo-specific references:

docs/commands/pr.md

@@ -20,6 +20,12 @@ f pr feedback
 # Pull feedback for specific PR and store as local todos
 f pr feedback 8 --todo
 f pr feedback https://github.com/owner/repo/pull/8 --todo
+
+# Full inline diff/review context is now the default
+f pr feedback 8
+
+# Use terse terminal output if needed
+f pr feedback 8 --compact
 ```
 
 ## Feedback Workflow
@@ -32,12 +38,27 @@ f pr feedback https://github.com/owner/repo/pull/8 --todo
 
 It then:
 
-1. Prints a concise actionable list in terminal.
-2. Writes a snapshot to `.ai/reviews/pr-feedback-<pr>.md`.
-3. Optionally (`--todo`) records feedback into `.ai/todos/todos.json` with dedupe via external refs.
+1. Prints an actionable list in terminal with inline review state and diff hunk context by default.
+2. Writes a markdown snapshot to `.ai/reviews/pr-feedback-<pr>.md`.
+3. Writes a machine-readable JSON snapshot to `.ai/reviews/pr-feedback-<pr>.json`.
+4. Writes a human review plan to `~/plan/review/<repo>-pr-<pr>-feedback.md`.
+5. Writes a PR-local execution artifact to `~/plan/review/<repo>-pr-<pr>-review-rules.md`.
+6. Writes a Kit system prompt to `~/plan/review/<repo>-pr-<pr>-kit-system.md`.
+7. Optionally (`--todo`) records feedback into `.ai/todos/todos.json` with dedupe via external refs.
+
+The review plan includes ready-to-run `kit` commands for:
+
+- deterministic repo review via `kit review`
+- preventative lint/review rule synthesis from the fetched GitHub feedback set
+
+The generated `*-review-rules.md` artifact contains the per-item resolution loop,
+prompt template, required response sections, and Kit-upgrade decision order so
+the workflow can be reopened without extra instructions.
 
 ## Notes
 
 - If no selector is passed, Flow resolves the PR from the current branch.
 - `f pr feedback --todo` is safe to re-run; existing feedback refs are not duplicated.
+- `f pr feedback` is full-context by default. Use `--compact` if you want the older terse terminal view.
+- `f pr feedback` now emits a Kit-ready handoff, so the same feedback set can drive a future review bot instead of staying trapped in GitHub comments.
 - `f pr open edit` remains the quickest path to tweak PR title/body from local editor.