fix: make global task autocomplete work by calling shell function instead of ./task

bin/setup-completion

@@ -78,7 +78,7 @@ output_completion_script() {
 # Taskfile completion for fish
 
 function __task_commands
-    ./task help 2>/dev/null | grep -A100 -e '---' | awk '{print $2}' | sed 's/^[[:space:]]*//' | grep -v '^$'
+    task help 2>/dev/null | grep -A100 -e '---' | awk '{print $2}' | sed 's/^[[:space:]]*//' | grep -v '^$'
 end
 
 # First argument position: offer the list of task commands.
@@ -100,7 +100,7 @@ _task_completions() {
     # $CURRENT is zsh's 1-indexed equivalent of bash's COMP_CWORD.
     # CURRENT == 2 means we're completing the first argument after `task`/`t`.
     if [[ $CURRENT -eq 2 ]]; then
-        IFS=$'\n' reply=($(./task help | grep -A999 -e '---' | awk '{print $2}' | sed 's/^[[:space:]]*//' | grep -v '^$'))
+        IFS=$'\n' reply=($(task help | grep -A999 -e '---' | awk '{print $2}' | sed 's/^[[:space:]]*//' | grep -v '^$'))
         IFS=$si
         if (( ${#reply} > 0 )); then
             _describe 'task' reply
@@ -125,7 +125,7 @@ _task_completions() {
     prev="${COMP_WORDS[COMP_CWORD-1]}"
 
     # compgen -W wants a single space-separated string, so tr is fine here.
-    commands=$(./task help | grep -A999 -e '---' | awk '{print $2}' | sed 's/^[[:space:]]*//' | grep -v '^$' | tr '\n' ' ')
+    commands=$(task help | grep -A999 -e '---' | awk '{print $2}' | sed 's/^[[:space:]]*//' | grep -v '^$' | tr '\n' ' ')
 
     if [[ ${COMP_CWORD} == 1 ]]; then
         COMPREPLY=( $(compgen -W "${commands}" -- "${cur}") )

bin/show-project-setup

@@ -1,6 +1,6 @@
 #!/usr/bin/env bash
 
-show-dotenv-setup() {
+show-project-setup() {
     TASK_DOTENV_DIR=${TASK_DOTENV_DIR:="$HOME/.config/taskfile.build"}
 
     local project_name
@@ -44,6 +44,6 @@ show-dotenv-setup() {
         echo "Age key file: not found"
     fi
 }
-sds() {
-    show-dotenv-setup "$@"
+sps() {
+    show-project-setup "$@"
 }

docs/library.md

@@ -85,15 +85,6 @@ rows+=("open-with-os|<path>|Open file or url with default application.")
 rows+=("open-url|<url>|Open url.")
 ```
 
-### show-dotenv-setup
-
-```bash
-rows+=("show-dotenv-setup||Show info about the dotenv setup.")
-```
-
-Displays the dotenv, KeePass / Pass entries, age key file, age public key,
-`.sops.yaml` config and `SOPS_AGE_KEY` status.
-
 ### setup-completion
 
 ```bash
@@ -279,6 +270,15 @@ rows+=("list-agekeys||List age keys from the default key file.")
 
 Lists the public keys, names and creation dates from `~/.config/sops/age/keys.txt`.
 
+### show-project-setup
+
+```bash
+rows+=("show-project-setup||Show project setup info.")
+```
+
+Displays the project name, dotenv, KeePass / Pass entries, age key file, age public key,
+`.sops.yaml` config and `SOPS_AGE_KEY` status.
+
 ### rename-dotenv
 
 ```bash

docs/sops.md

@@ -26,7 +26,7 @@ task init-dotenv-sops taskfile.build
 Show the current setup with:
 
 ```bash
-task show-dotenv-setup
+task show-project-setup
 ```
 
 List age keys in the default key file:

prompt.md

@@ -1,15 +1,15 @@
 ---
-title: "Prompt-log improvements"
+title: "Make global autocomplete work"
 author: "Janik von Rotz <login@janikvonrotz.ch>"
-state: completed
-date_completed: 2026-10-13
+state: review
+date_completed:
 harness: pi
 model: moonshotai/Kimi-K2.6
-input_tokens: 36301
-output_tokens: 11701
+input_tokens:
+output_tokens:
 ---
 
-# Prompt-log improvements
+# Make global autocomplete work
 
 Note: @Clanker refers to the "ai agent" (you) who is working on this prompt file.
 
@@ -25,28 +25,38 @@ Note: @Clanker refers to the "ai agent" (you) who is working on this prompt file
 
 ## Task
 
-Here are some improvements for the work with `prompt-log`.
+If I try to run the zsh `task` autocomplete in a folder without `task` file. Then the autocomplete does not work:
 
-When I run `create-prompt` and there is already a `prompt.md` I want you to run `commit-prompt` in order to store the current prompt file.
-
-This prompt file has been updated with new instructions. When the task is completed the state of the prompt file must be `review`. The clanker then asks the user if the result checks. The worklog is used whether the author confirms or not. I added this text to the prompt.md:
-
-```markdown
-@Clanker When task is completed set frontmatter to review. Prompt the author to verify the results.
-
-## Worklog
-
-@Clanker Add a summary here once the task has been completed.
-
-@Clanker If author confirms results set frontmatter state to completed, otherwise to aborted. Update date, harness and model. If you have access to session info also add token count.
+```
+task <tab> ini_task_completions:7: no such file or directory: ./task
+_task_completions:7: no such file or directory: ./task
+_task_completions:7: no such file or directory: ./task
 ```
 
-Check the text for typos and logical issues. Then update `prompt.md.template`.
+The `task` command is function loaded into the profile
 
-@Clanker When task is completed set frontmatter to review. Prompt the author to verify the results.
+```bash
+[~]$ which task
+task () {
+	if [[ -f "./task" && -x "./task" ]]
+	then
+		./task "$@"
+	else
+		"$HOME/.local/bin/task" "$@"
+	fi
+}
+```
+
+See `docs/usage.md` for details.
+
+@Clanker When task is completed set frontmatter state to review. Prompt the author to verify the results.
 
 ## Worklog
 
-Consolidated Worklog and Backlog sections into a single Worklog section. Removed the Backlog section from both `prompt.md` and `prompt.md.template`. The Worklog now handles all outcomes: add a summary upon task completion, set frontmatter state to `completed` if author confirms or `aborted` if not, and update metadata accordingly. Also fixed typos in `prompt.md` ("fiel" → "file", "upated" → "updated", "to state" → "the state", "th results" → "the results").
+The shell completion scripts in `bin/setup-completion` were calling `./task help` directly instead of the user's `task()` shell function. When no local `./task` file existed, this produced `_task_completions:7: no such file or directory: ./task`.
 
-Updated `bin/create-prompt` to automatically run `commit-prompt` when the target prompt file already exists, ensuring the current prompt is stored before a new one is created. Verified `prompt.md.template` contains the new instructions for setting state to `review` upon task completion and handling author confirmation in the Worklog.
+Fixed by replacing `./task help` with `task help` in all three completion scripts (bash, zsh, fish). This lets the shell function resolve the correct task binary — local `./task` if present, otherwise `$HOME/.local/bin/task` — making completions work in any directory.
+
+File changed: `bin/setup-completion`.
+
+@Clanker If author confirms results set frontmatter state to completed, otherwise to aborted. Update date, harness and model. If you have access to session info also add token count.

task.template

@@ -18,7 +18,6 @@ help-table() {
     rows+=("template-dotenv||Generate .env from .env.template.")
     rows+=("open-with-os|<path>|Open file or url with default application.")
     rows+=("open-url|<url>|Open url.")
-    rows+=("show-dotenv-setup||Show info about the dotenv setup.")
     rows+=("setup-completion|[shell]|Show shell completion instructions.")
     rows+=("list-aliases||List all library command aliases.")
     rows+=("check-help-width||Show help lines that exceed terminal width.")
@@ -43,6 +42,7 @@ help-table() {
     rows+=("load-dotenv|<env>|Load and export .env file.")
     rows+=("init-dotenv-sops||Initialize age key and sops config for the project.")
     rows+=("list-agekeys||List age keys from the default key file.")
+    rows+=("show-project-setup||Show project setup info.")
     rows+=("rename-dotenv|<env><env>|Rename .env file.")
     rows+=("remove-dotenv|<env>|Remove .env file.")
     rows+=("backup-dotenv-files|<path>|Archive and copy .env files to target location.")