Hijack agent-shell-resume-session to use the session picker

README.org

@@ -20,12 +20,13 @@ A soft fork of [[https://github.com/xenodium/agent-shell][agent-shell]] with ext
 - Append-in-place rendering for streamed fragments: tokens append to existing fragment bodies without rebuilding, with boundary-newline normalization so paragraph-break chunks don't compound newlines and an empty =agent_message_chunk= mid-stream is rewritten to a paragraph break so two content blocks in the same turn don't run together ([[https://github.com/timvisher-dd/agent-shell-plus/pull/7][#7]])
 - DWIM context insertion: inserted context lands at the prompt and fragment updates no longer drag process-mark past it ([[https://github.com/timvisher-dd/agent-shell-plus/pull/7][#7]])
 - Runtime buffer invariant checking with event tracing and violation debug bundles, including head + tail snapshots for long buffers ([[https://github.com/timvisher-dd/agent-shell-plus/pull/7][#7]])
-- Render =session/update= chunks streamed after =session/prompt= resolves so a Claude Code Stop-hook bounce-and-regen turn no longer freezes the buffer mid-conversation
-- Surface raw claude-agent-acp SDK messages (including hook lifecycle events) in the debug log when =agent-shell-logging-enabled= is set, so Stop-hook =decision:block= cycles and other hook-driven turn behavior are visible
-- Tunable markdown-overlay debounce via =agent-shell-markdown-overlay-debounce-delay= (default 0.15s) for slow terminals or streaming-debug sessions
-- Bug fix for upstream =shell-maker-define-major-mode= mode-map quoting — without it, every =agent-shell-mode= invocation emits =void-function keymap= because the bare keymap value gets spliced into a backquote that re-evaluates =(keymap ...)= as a function call (worth upstreaming separately)
-- Live-validate workflow doc (=.agents/commands/live-validate.md=) describing the batch-mode rendering verification used for rendering-pipeline changes
-- =gfm-mode= compose buffer for the interactive =agent-shell-queue-request=, replacing the read-string minibuffer prompt (non-interactive callers still pass =PROMPT= directly)
+- Render =session/update= chunks streamed after =session/prompt= resolves so a Claude Code Stop-hook bounce-and-regen turn no longer freezes the buffer mid-conversation ([[https://github.com/timvisher-dd/agent-shell-plus/pull/7][#7]])
+- Surface raw claude-agent-acp SDK messages (including hook lifecycle events) in the debug log when =agent-shell-logging-enabled= is set, so Stop-hook =decision:block= cycles and other hook-driven turn behavior are visible ([[https://github.com/timvisher-dd/agent-shell-plus/pull/7][#7]])
+- Tunable markdown-overlay debounce via =agent-shell-markdown-overlay-debounce-delay= (default 0.15s) for slow terminals or streaming-debug sessions ([[https://github.com/timvisher-dd/agent-shell-plus/pull/7][#7]])
+- Bug fix for upstream =shell-maker-define-major-mode= mode-map quoting — without it, every =agent-shell-mode= invocation emits =void-function keymap= because the bare keymap value gets spliced into a backquote that re-evaluates =(keymap ...)= as a function call (worth upstreaming separately) ([[https://github.com/timvisher-dd/agent-shell-plus/pull/7][#7]])
+- Live-validate workflow doc (=.agents/commands/live-validate.md=) describing the batch-mode rendering verification used for rendering-pipeline changes ([[https://github.com/timvisher-dd/agent-shell-plus/pull/7][#7]])
+- =gfm-mode= compose buffer for the interactive =agent-shell-queue-request=, replacing the read-string minibuffer prompt (non-interactive callers still pass =PROMPT= directly) ([[https://github.com/timvisher-dd/agent-shell-plus/pull/9][#9]])
+- =agent-shell-resume-session= repurposed from an ID-prompt command into a session-picker entry point: opens the same fuzzy-completion picker that =agent-shell-session-strategy= ='prompt= uses, and force-shows session IDs in the picker (via a new =:show-session-id= keyword on =agent-shell--start=) so users with an ID in hand can find it via their completion framework ([[https://github.com/timvisher-dd/agent-shell-plus/pull/11][#11]])
 
 -----
 

agent-shell.el

@@ -1180,19 +1180,20 @@ Works from both shell and viewport buffers."
         (agent-shell--display-buffer new-shell-buffer)))))
 
 ;;;###autoload
-(defun agent-shell-resume-session (session-id)
-  "Resume an existing agent session by SESSION-ID.
-
-Prompts for agent selection and starts a new shell that resumes
-the session identified by SESSION-ID."
-  (interactive "sSession ID: ")
-  (when (string-empty-p (string-trim session-id))
-    (user-error "Session ID cannot be empty"))
+(defun agent-shell-resume-session ()
+  "Resume an existing agent session via the session picker.
+
+Prompts for agent selection, then starts a new shell that lists known
+sessions and lets the user pick one to resume.  Session IDs are shown
+in the picker (overriding `agent-shell-show-session-id' for this shell)
+so users who know an ID can find it via their completion framework."
+  (interactive)
   (agent-shell--start :config (or (agent-shell--resolve-preferred-config)
                                   (agent-shell-select-config
                                    :prompt "Resume with agent: ")
                                   (error "No agent config found"))
-                      :session-id session-id
+                      :session-strategy 'prompt
+                      :show-session-id t
                       :new-session t))
 
 ;;;###autoload
@@ -2815,14 +2816,15 @@ FUNCTION should be a function accepting keyword arguments (&key ...)."
                    (list (car pair) (cdr pair)))
                  alist)))
 
-(cl-defun agent-shell--start (&key config no-focus new-session session-strategy session-id fork-session-id outgoing-request-decorator)
+(cl-defun agent-shell--start (&key config no-focus new-session session-strategy show-session-id session-id fork-session-id outgoing-request-decorator)
   "Programmatically start shell with CONFIG.
 
 See `agent-shell-make-agent-config' for config format.
 
 Set NO-FOCUS to start in background.
 Set NEW-SESSION to start a separate new session.
 SESSION-STRATEGY overrides `agent-shell-session-strategy' buffer-locally.
+SHOW-SESSION-ID overrides `agent-shell-show-session-id' buffer-locally.
 SESSION-ID resumes an existing session by its id string.
 FORK-SESSION-ID forks an existing session by its id string.
 OUTGOING-REQUEST-DECORATOR is passed through to `acp-make-client'."
@@ -2922,6 +2924,8 @@ variable (see makunbound)"))
         (map-put! agent-shell--state :fork-session-id fork-session-id))
       (when session-strategy
         (setq-local agent-shell-session-strategy session-strategy))
+      (when show-session-id
+        (setq-local agent-shell-show-session-id show-session-id))
       ;; Show deferred welcome text,
       ;; but first wipe buffer content.
       (let ((inhibit-read-only t))