list-work-items

Invocation: User- and model-invocable
Argument hint: [filter description]
Allowed tools: Bash(${CLAUDE_PLUGIN_ROOT}/scripts/config-*), Bash(${CLAUDE_PLUGIN_ROOT}/skills/work/scripts/*)
Source: skills/work/list-work-items/SKILL.md

List Work Items

!${CLAUDE_PLUGIN_ROOT}/scripts/config-read-context.sh !${CLAUDE_PLUGIN_ROOT}/scripts/config-read-skill-context.sh list-work-items !${CLAUDE_PLUGIN_ROOT}/scripts/config-read-agents.sh

If no “Agent Names” section appears above, use these defaults: accelerator:reviewer, accelerator:codebase-locator, accelerator:codebase-analyser, accelerator:codebase-pattern-finder, accelerator:documents-locator, accelerator:documents-analyser, accelerator:web-search-researcher.

Work items directory: !${CLAUDE_PLUGIN_ROOT}/scripts/config-read-path.sh work Work item ID pattern: !${CLAUDE_PLUGIN_ROOT}/scripts/config-read-work.sh id_pattern Default project code: !${CLAUDE_PLUGIN_ROOT}/scripts/config-read-work.sh default_project_code Active integration: !${CLAUDE_PLUGIN_ROOT}/scripts/config-read-work.sh integration

The Active integration line above gates the sync-status rendering (Step 4). config-read-work.sh integration exits 0 with an empty line when no integration is configured, so branch on the string, not the exit code: treat a non-empty value as integration configured and an empty value as not configured.

Work Item Template

The following template defines the work item schema and field defaults. Hint values for filter parsing are extracted at runtime via work-item-template-field-hints.sh.

!${CLAUDE_PLUGIN_ROOT}/scripts/config-read-template.sh work-item

You are tasked with listing and filtering work items from the configured work items directory. This is a read-only skill — never write any files and never spawn sub-agents. The entire flow uses filesystem reads and the companion scripts listed in allowed-tools.

Step 1: Resolve Filter

If an argument was provided, parse it as a filter expression using the following precedence rules. The first rule that matches wins.

Before applying rules 3–4, call work-item-template-field-hints.sh for each of kind, status, and priority to populate the known template-comment values:

${CLAUDE_PLUGIN_ROOT}/skills/work/scripts/work-item-template-field-hints.sh kind
${CLAUDE_PLUGIN_ROOT}/skills/work/scripts/work-item-template-field-hints.sh status
${CLAUDE_PLUGIN_ROOT}/skills/work/scripts/work-item-template-field-hints.sh priority

Each call outputs one value per line. Collect these into three sets: known kinds, known statuses, and known priorities. These hints inform the shorthand rules below but do not restrict what values may appear on work items — legacy values like todo, done, or adr-creation-task are matchable only via the explicit structured form (rule 2).

Filter Precedence Rules

Rule 1 — Presentation keywords (changes rendering, not filtering): hierarchy, as a tree, show hierarchy. These may combine with a filter (e.g. hierarchy under 0042). Strip the keyword from the argument and continue parsing the remainder, if any, through rules 2–5.

Rule 2 — Explicit structured forms (keyword identifies the field):

• tagged <value> or with tag <value> → filter by tag
• under <value> or children of <value> → filter by parent
• status <value> → filter by status (matches any value on any work item)
• kind <value> → filter by kind (matches any value on any work item)
• priority <value> → filter by priority
• about <text> → free-text title search (case-insensitive substring)

Rule 3 — Multi-token template-value shorthand: two or more tokens that each match a known template-comment value in different fields. For example, bugs in review → kind: bug AND status: review (after singularising bugs to bug and recognising in-progress, review, etc. as status values with filler words like in, only, all stripped). Both tokens must match values from distinct fields. If either token is ambiguous across fields, fall through to rule 5.

Rule 4 — Single-token template-value shorthand: one token matching a known kind, status, or priority value from the template comments. Map common plurals (bugs→bug, epics→epic, stories→story, tasks→task, spikes→spike) and common synonyms (drafts→draft). If the token matches values in more than one field, ask the user for disambiguation rather than guessing.

Rule 5 — Free-text title search: anything that does not match rules 1–4 is treated as a case-insensitive substring search against the title: frontmatter field.

Always echo the interpreted filter before showing results so the user can rephrase if the parse was wrong. Example: Filter: status=draft (3 matches).

If no argument was provided: filter is “all work items, no filter”.

Step 2: Scan Work Items Directory

1. Check that {work_dir} exists. If not, print:

   Work items directory `{work_dir}` not found.
   Check the `paths.work` configuration or run `/create-work-item` to
   create the first work item.

   and exit cleanly.

2. List all *.md files in {work_dir}. The discovery glob is broadened from a literal numeric prefix to *.md because the work-item ID pattern is configurable (work.id_pattern) and legacy NNNN-*.md files may coexist with project-coded files (PROJ-NNNN-*.md) during a pattern transition. A file is treated as a work item iff work-item-common.sh:wip_is_work_item_file returns success — that is, the file has YAML frontmatter and a non-empty id field (or work_item_id on legacy files). Files lacking either are silently excluded; files with malformed frontmatter emit a one-line warning to stderr and are skipped.

3. If no work-item files exist after filtering, print:

   No work items found in `{work_dir}`.

   and exit cleanly.

4. Extract frontmatter from all work items in a single pass. Work item directories can contain dozens of files, so reading each one individually would be too slow. Instead, use a single Bash command to extract the frontmatter fields from every matching file at once.

   Example approach — run one awk command across all *.md files:

   for f in {work_dir}/*.md; do
     # Filter to true work items via wip_is_work_item_file
     if ! bash -c "source ${CLAUDE_PLUGIN_ROOT}/skills/work/scripts/work-item-common.sh && wip_is_work_item_file '$f'"; then
       continue
     fi
     awk -v file="$f" '
       NR==1 && /^---[[:space:]]*$/ { in_fm=1; next }
       NR==1 { print file "\tERROR\tno frontmatter"; exit }
       in_fm && /^---[[:space:]]*$/ { closed=1; exit }
       in_fm { print file "\t" $0 }
       END { if (in_fm && !closed) print file "\tERROR\tunclosed frontmatter" }
     ' "$f"
   done

   This outputs one line per frontmatter field per file (tab-delimited: filepath, field line) plus ERROR lines for malformed files. Parse the output to build the work item list in memory.

   • Lines containing ERROR no frontmatter: warn "<filename>: skipped — no frontmatter" and exclude the file.
   • Lines containing ERROR unclosed frontmatter: warn "<filename>: skipped — unclosed frontmatter" and exclude.
   • For each valid file, derive the work item ID from the filename via wip_extract_id_from_filename. The compiled scan regex respects the configured work.id_pattern; a legacy 0042-foo.md file under a {project}-{number:04d} pattern is matched by a legacy-fallback path so the file remains visible in the listing. The filename prefix remains the authoritative work item ID, even if the id field (or work_item_id on legacy files) in frontmatter differs.

5. Mixed-prefix discoverability hint: when the listing detects both files matching the legacy [0-9]{4}- shape AND files matching the configured {project} pattern in the same corpus, prepend a single informational line to the output before the table:

   note: mixed prefix corpus detected — N legacy items, M project-prefixed
   items. Run /accelerator:migrate to normalise.

   The note appears once per invocation (not per file) and is suppressed when the configured pattern lacks {project}.

6. From the extracted frontmatter lines, parse these fields for each work item (all optional — missing fields are recorded as absent, not as errors):

   • title — the human-readable title
   • kind — the work item kind
   • status — the current status
   • priority — the priority level
   • tags — a YAML inline array (e.g. [backend, api])
   • parent — the parent work item number
   • external_id — the remote tracker’s identifier, used only for the sync-status label in Step 4. It is already present in the single-pass frontmatter stream above (the awk prints every frontmatter line), so reading it adds no extra per-file process. Read it only when an integration is configured; otherwise ignore it.

   Sync classification (only when an integration is configured). For each file that passed the frontmatter validity check above, classify it presence-based:

   • external_id absent → unsynced (never an error — the item was simply never pushed).
   • external_id present but normalising to empty (after stripping surrounding quotes and whitespace — e.g. external_id: "") → unsynced.
   • external_id present and non-empty after normalisation → synced.

   Do not hand-roll the normalisation: pass the raw external_id value (or nothing, when the field is absent) to the authoritative classifier work-item-sync-label.sh (Step 4), which owns the trimming rule and the label vocabulary. A file that Step 2 already flagged as malformed stays a skip, not an unsynced row. The filename remains the authoritative displayed ID — external_id never changes the displayed ID.

   This presence-only synced/unsynced split is the floor. When a last-sync.json baseline exists, Step 4 upgrades tracked items (those with an external_id and a baseline entry) to the three baseline-dependent states — locally-modified, remotely-modified, conflict — via a single bulk remote read and the shared change-detection engine, degrading back to this presence-only floor whenever the remote is unreachable or no baseline exists. See “Sync Status Labels” in Step 4.

Step 3: Apply Filter

Apply the parsed filter from Step 1 to the scanned work items.

• “All, no filter”: keep every work item.
• Status/kind/priority filter: match the field value exactly (case-sensitive, matching the raw frontmatter value). Work items missing the filtered field are excluded from the result (not errors).
• Tag filter: parse the raw tags value (e.g. [backend, api]) into individual tag strings. A work item matches if any tag equals the filter value. Work items with tags: [], empty tags:, or absent tags field do not match (and are not errors).
• Parent filter (under X): normalise both the filter value and each work item’s parent field via work-item-common.sh:wip_canonicalise_id before comparison. The canonicaliser strips quotes, accepts short and long forms, and zero-pads to the configured pattern’s width (or pre-pends the default project code when the pattern requires {project} and the input is a bare number). So under default config parent: "0042", parent: 0042, parent: 42, and parent: "42" all match under 0042 or under 42. Under {project}-{number:04d} config with default_project_code: PROJ, parent: "PROJ-0042", parent: "0042" (legacy), and parent: 42 all canonicalise to PROJ-0042 and match under PROJ-0042 or under 42.
• Free-text title search (about X or rule 5): case-insensitive substring match against the title: frontmatter value. Work items without a title field are excluded.
• Combined filters (rule 3): all conditions must hold (AND).

Step 4: Render

Sync Status Labels (only when an integration is configured)

When the Active integration read at the top of this skill is a non-empty string, each rendered work item carries a sync-status label. When it is empty, render exactly as today — no Sync column, no label, output unchanged.

All five label states are owned by one source of truth so the table and hierarchy views never drift:

${CLAUDE_PLUGIN_ROOT}/skills/work/scripts/work-item-sync-label.sh --label <state>

where <state> ∈ synced, unsynced, locally-modified, remotely-modified, conflict renders the markdown-native label (🟢 synced, ⚪ unsynced, 🔵 locally modified, 🟣 remotely modified, 🔴 conflict). Every pair differs in both glyph and text, and the labels are markdown-native (a Unicode glyph + text) emitted into the conversation’s table — never ANSI escape codes, which would surface as literal \033[…] text.

Which states are reachable depends on the baseline

Resolve the baseline path:

${CLAUDE_PLUGIN_ROOT}/skills/work/scripts/work-item-sync-baseline.sh path

• No last-sync.json baseline file (the resolved path does not exist): there is no referent for change detection, so classify every item presence-only — external_id non-empty → synced, else unsynced (pass the raw value to work-item-sync-label.sh <external-id-value>, which owns the trimming rule). Do no remote read. This is exactly the 0047 behaviour.

• Baseline exists: render the full five-state set for tracked items, driven by a SINGLE bulk remote read plus the shared engine. Items with no external_id, or no baseline entry, stay presence-only.

Five-state classification (baseline present)

1. Bulk pre-filter read — one call, never N. Collect the non-empty external_ids of all listed items and fetch their remote updated stamps in one call:

   ${CLAUDE_PLUGIN_ROOT}/skills/work/scripts/work-item-fetch-remote.sh \
     --integration <sys> search --keys <comma-separated external_ids>

   It returns { "found": {\<key>:{updated}}, "absent": [...], "indeterminate": [...] } — the adapter chose the per-tracker strategy, so you never branch on tracker. Graceful degradation: if the bridge exits non-zero (remote unreachable / timed out), do not retry or hang — fall back to presence-only for every item, render the listing, and exit 0. One bulk call bounds the whole degradation path; a key that lands in indeterminate is likewise treated as unknown, never as remote-absent.

2. Per item, derive the remote status to hand the engine:

   • no external_id → presence-only unsynced (skip the engine).
   • external_id present but no baseline entry (work-item-sync-baseline.sh get <id> prints nothing) → presence-only (synced); skip the engine.
   • key in indeterminate, or the bulk read degraded → --remote-status indeterminate.
   • key in absent → --remote-status absent.
   • key in found → --remote-status present --remote-updated <its updated>. If that updated differs from the baseline entry’s remote_updated_at, this item is in the genuinely-changed minority: fetch its body (work-item-fetch-remote.sh … show --external-id <key>), project + canonicalise it to the comparable local shape (jira: .fields.summary → title, .fields.description ADF through jq -S; linear: .data.issue.title + .data.issue.description Markdown, no jq -S), write it to a temp file, and pass --remote-body-file. Reserve show for this minority and emit classifying item k of N progress so a long pass never reads as a hang.

3. Classify via the shared engine and render its keyword:

   ${CLAUDE_PLUGIN_ROOT}/skills/work/scripts/work-item-sync-classify.sh \
     --file <path> --external-id <key> \
     --baseline "$(work-item-sync-baseline.sh get <id>)" \
     --timestamp "$(jq -r '.timestamp // 0' "$(work-item-sync-baseline.sh path)")" \
     --remote-status <present|absent|indeterminate> \
     [--remote-updated <iso>] [--remote-body-file <tmp>]

   The engine prints one of synced | unsynced | locally-modified | remotely-modified | conflict | remote-absent | indeterminate. Feed the first five straight to work-item-sync-label.sh --label <state>. For remote-absent and indeterminate, render the presence-only label (synced, since the item carries an external_id): the remote state is unknown or the issue is gone, and the listing must never fail or hang on it.

The canonical-tree-fence example below stays label-free.

Default Rendering (table)

Present the filtered work items as a markdown table with these columns:

| ID | Title | Kind | Status | Priority |

When an integration is configured (see “Sync Status Labels” above), append a Sync column carrying each item’s label:

| ID | Title | Kind | Status | Priority | Sync |

• Sort rows by work item number ascending.
• Render missing fields as —.
• If a column would be — for every row in the current result set, suppress that column entirely to reduce noise. For example, a listing of only legacy work items (which lack priority) would omit the Priority column. The Sync column follows the same rule in reverse: it is present only when an integration is configured, and suppressed entirely when one is not (there is no label to show).

Hierarchy Rendering

If a hierarchy presentation keyword was detected in Step 1:

• Work items with no parent (or empty parent) appear at the top level.
• Work items whose parent points to a work item in the current result set are rendered as children. Each parent→children group prints as a tree using Unicode box-drawing characters. Children use ├── for all entries except the last in the group, which uses └── . Indent two spaces per depth level. Example:

<!— canonical-tree-fence —> NNNN — parent title (kind: <kind>, status: <status>) ├── NNNN — child 1 title (kind: <kind>, status: <status>) ├── NNNN — child 2 title (kind: <kind>, status: <status>) └── NNNN — last child title (kind: <kind>, status: <status>) <!— /canonical-tree-fence —>

No ASCII fallback is attempted; terminals without Unicode support will render mojibake. Users on such terminals can re-display the hierarchy via /list-work-items.

• Sync labels in the tree (only when an integration is configured): append each item’s sync label — obtained from work-item-sync-label.sh, the same source of truth the table uses — to the end of its line, after the (kind: …, status: …) segment. For example:

  0042 — User Auth Rework (kind: epic, status: ready) 🟢 synced
    └── 0043 — Login form (kind: story, status: draft) ⚪ unsynced

  The shared canonical-tree-fence example above is deliberately kept label-free: it is an integration-agnostic static example shared byte-for-byte with /refine-work-item (which has no integration gate), so the sync label is applied to the live rendered output, never to that example.

• Work items whose parent points to a work item number that does not exist in the result set appear at the top level with a suffix: (parent NNNN not found).

• Cycle detection: before rendering, walk the parent chain for each work item. If a work item is visited twice during a walk, it is part of a cycle. Render all cyclic work items at the top level with a (cycle) marker. This ensures bounded execution — no infinite loops.

Empty Results

• If zero work items match the filter, print:

  No work items matched: <filter description>.

  Do not render an empty table. If the active filter was a free-text title search (rule 5), append:

  Tip: to filter by field value, use `status <value>`, `kind <value>`,
  or `tagged <value>`.

Filter Echo

Always print the interpreted filter and match count above the table:

Filter: status=draft (3 matches)

or for a parent filter:

Children of 0042 (2 matches)

or for no filter:

All work items (29 total)

Quality Guidelines

• Read-only: never write any files. This skill only reads and displays.
• No sub-agents: never spawn sub-agents. All work is done via filesystem reads and companion scripts.
• No hardcoded field values: never assume a specific set of status, kind, or priority values. The template’s comments list shipping defaults, not a closed set. Users may override the template with custom values.
• Explicit structured filters are universal: status <value>, kind <value>, etc. match any value present on any work item, not just template defaults. This is how legacy values like todo or adr-creation-task are reachable.
• Resilient to malformed work items: missing or unclosed frontmatter must not crash the listing — warn using the resolved filename and continue. Warning phrasing should match work-item-read-field.sh: “no frontmatter” / “unclosed frontmatter”.
• Filename is authoritative: the ID extracted from the filename (via wip_extract_id_from_filename) is the work item ID, even if the id field (or work_item_id on legacy files) in frontmatter differs. This applies to both legacy bare-number filenames and project-coded filenames.
• Hierarchy safety: hierarchy rendering must terminate in bounded time even if parent cycles exist. Detect cycles and render affected work items flat with a marker.

!${CLAUDE_PLUGIN_ROOT}/scripts/config-read-skill-instructions.sh list-work-items