Skill: Trello Load

Description

Connects to Trello via the MCP server, pulls all cards from the Sprint_Backlog list, and creates well-formed beans directly from each card's content. Beans are created with Approved status (ready for work). After each bean is created, the source card is moved to the In_Progress list on Trello. This runs fully non-interactively — no user prompts, no clarifying questions — suitable for cron jobs or automated pipelines like /long-run.

Trigger

• •Invoked by the /trello-load slash command.
• •Requires the Trello MCP server to be configured and accessible.
• •Requires ai/beans/_index.md and ai/beans/_bean-template.md to exist.

Inputs

Process

Phase 1: Board Selection (Automatic)

1. •

   Check Trello connection — Call mcp__trello__get_health to verify the MCP server is responsive. If unhealthy, report the error and stop.

2. •

   Determine board — Three strategies, in order: a. If --board <id> was provided, use that board ID directly. b. Call mcp__trello__list_boards and look for a board whose name matches the current project directory name (case-insensitive). For this repo, that means a board named "Foundry". c. If no match, use the first board returned. Log which board was selected and why.

3. •

   Set active board — Call mcp__trello__set_active_board with the resolved board ID.

Phase 2: List Discovery

1. •

   Fetch lists — Call mcp__trello__get_lists for the selected board.

2. •

   Find Sprint_Backlog list — Search the returned lists for one matching "Sprint_Backlog" using flexible matching:

   • •Normalize both the target name and each list name by:
     • •Converting to lowercase
     • •Replacing underscores, hyphens, and spaces with a single common separator
   • •Match if the normalized names are equal
   • •Examples that match: "Sprint_Backlog", "sprint_backlog", "Sprint Backlog", "sprint-backlog", "SPRINT_BACKLOG"
   • •If no match found, show available list names and stop with an error.
3. •

   Find In_Progress list — Same flexible matching for "In_Progress":

   • •Matches: "In_Progress", "in_progress", "In Progress", "in-progress", "IN_PROGRESS"
   • •If no match found, show available list names and stop with an error.
4. •

   Report discovered lists — Confirm to the user:

   code

   Found lists:
   - Sprint Backlog: "[actual list name]" (N cards)
   - In Progress: "[actual list name]"
   

Phase 3: Card Loading

1. •

   Fetch cards — Call mcp__trello__get_cards_by_list_id with the Sprint_Backlog list ID.

2. •

   Check for empty backlog — If no cards are returned, inform the user and stop:

   code

   Sprint_Backlog is empty — nothing to process.
   

3. •

   Fetch full card details — For each card, gather complete data: a. Card details — Call mcp__trello__get_card with includeMarkdown: true to get name and description. b. Checklists — For each checklist on the card, call mcp__trello__get_checklist_items to get all checklist items and their completion state. c. Comments — Call mcp__trello__get_card_comments to get all comments.

4. •

   Compile card text — For each card, assemble a single text block:

   code

   # [Card Name]
   
   ## Description
   [Card description text]
   
   ## Checklists
   ### [Checklist Name]
   - [x] Completed item
   - [ ] Incomplete item
   ...
   
   ## Comments
   - [Author] ([date]): [comment text]
   ...
   

   Omit sections that are empty (e.g., if no checklists, skip that section).

5. •

   Log summary — Show what will be processed:

   code

   Found N cards in Sprint_Backlog:
   1. [Card Name] — [first 80 chars of description]
   2. [Card Name] — [first 80 chars of description]
   ...
   
   Creating beans and moving cards to In_Progress.
   

6. •

   Handle dry run — If dry_run is true, show the compiled text and proposed bean fields for each card and stop. Do not create beans or move any cards.

Phase 4: Sequential Bean Creation (Non-Interactive)

1. •

   Process each card — For each card in order:

   a. Announce — Log which card is being processed:

   code

   --- Processing card 1/N: [Card Name] ---
   

   b. Re-read ai/beans/_index.md to get the current max bean ID. Another agent may have added beans since the last read.

   c. Assign the next sequential ID (max + 1).

   d. Create the bean directly — Build a well-formed bean from the card content. Do NOT invoke /backlog-refinement — create the bean inline to avoid interactive dialogue. Map card fields to bean fields:

   • •Title: Use the card name. Generate a slug from it.
   • •Problem Statement: Derive from the card description. If the description is empty, expand the card name into a problem statement (e.g., "Create a test that..." → "The project lacks an end-to-end test that...").
   • •Goal: Infer the desired outcome from the description and any checklists.
   • •Scope — In Scope: Extract from checklist items and description. If sparse, infer reasonable scope from the title.
   • •Scope — Out of Scope: Use best judgment to set boundaries.
   • •Acceptance Criteria: Convert checklist items to acceptance criteria. Always include the standard "All tests pass" and "Lint clean" criteria. Ensure at least 3 total criteria.
   • •Priority: Default to High (sprint backlog items are prioritized). If the card has labels, use them as hints (red=High, yellow=Medium, green=Low).
   • •Status: Approved (not Unapproved — these come from a curated sprint backlog and are pre-approved).
   • •Category: Infer from content (App for code/test/UI changes, Process for workflow/agent changes, Infra for CI/CD/git changes).
   • •Notes: Include "Source: Trello card '[Card Name]'" and a link to the card URL.

   e. Write the bean — Create ai/beans/BEAN-NNN-<slug>/bean.md and append to _index.md immediately.

   f. Move card to In_Progress — Call mcp__trello__move_card with the card ID and the In_Progress list ID. Log the move:

   code

   Moved "[Card Name]" → In_Progress | Created BEAN-NNN
   

   g. Handle errors gracefully — If bean creation or card move fails:

   • •Log the error
   • •Skip the card, leave it in Sprint_Backlog
   • •Continue to the next card (do not stop)

Phase 5: Summary

1. •Present completion report — After all cards are processed (or processing is stopped early), show:
   code

   Trello Load Complete:
   - Cards processed: N/M
   - Cards moved to In_Progress: N
   - Cards skipped: N
   - Total beans created: N
   
   Created beans:
   | Bean ID | Title | Source Card |
   |---------|-------|-------------|
   | BEAN-NNN | [title] | [card name] |
   ...
   

Outputs

Quality Criteria

• •Every card's full context (name, description, checklists, comments) is used to populate the bean — no data is silently dropped.
• •List name matching is flexible and case-insensitive.
• •Cards are only moved to In_Progress after bean creation succeeds.
• •If bean creation fails for a card, it stays in Sprint_Backlog.
• •Beans are created with Approved status (pre-approved from sprint backlog).
• •No user interaction required — the entire flow runs autonomously.
• •Clear progress logging (card N/M) throughout processing.
• •The final summary accurately reflects what was created and moved.

Error Conditions

Dependencies

• •Trello MCP server (configured and accessible)
• •mcp__trello__list_boards, mcp__trello__get_lists, mcp__trello__get_cards_by_list_id
• •mcp__trello__get_card, mcp__trello__get_card_comments, mcp__trello__get_checklist_items
• •mcp__trello__move_card
• •Backlog index at ai/beans/_index.md
• •Bean template at ai/beans/_bean-template.md