Skill

Compile the most recent task-level comment for each active project in a configured Asana portfolio and save the result as a dated Markdown file to a local folder. The local folder may be a vanilla directory or one that a sync client (e.g., Google Drive for Desktop, Dropbox, OneDrive, Syncthing, rclone) mirrors elsewhere — this skill does no remote uploads of its own.

Before running this skill, edit the placeholders below. They are written in <ANGLE_BRACKETS> and listed in CONFIGURATION.md at the root of this package.

Fixed identifiers (edit these)

• Asana portfolio GID: <PORTFOLIO_GID> (name: <PORTFOLIO_NAME>)
• Output folder (absolute local path): <OUTPUT_FOLDER>
• Output filename pattern: <FILE_PREFIX>-Status-YYYY-MM-DD.md (today's local date)
• Filter field: Asana custom field named <WORKFLOW_CATEGORY_FIELD>
  • Excluded values (case-insensitive): <EXCLUDED_VALUE_1>, <EXCLUDED_VALUE_2>, ... plus empty/null
  • Treat all other values as "active"
• Optional responsible-person custom field: <RESPONSIBLE_FIELD> (falls back to project owner if unset)

Why this design

Many institutional Google Workspaces block third-party Drive API apps at the domain policy level, which prevents most "write to Drive" MCP connectors from working. Writing the file locally and letting a first-party sync client (Google Drive for Desktop, Dropbox, OneDrive, etc.) mirror it sidesteps that policy entirely. The whole workflow is local file I/O plus the Asana MCP — nothing else.

If you don't need sync at all, point <OUTPUT_FOLDER> at any local directory.

Required connectors / tools

• Asana MCP with at least: get_items_for_portfolio, get_tasks, get_task
• Local file system access via the agent's Read/Write tools or shell

If the Asana MCP is unavailable, stop and report.

Steps

1. Fetch and filter portfolio projects

Call get_items_for_portfolio:

• portfolio_gid = "<PORTFOLIO_GID>"
• limit = 100
• opt_fields = "name,gid,permalink_url,current_status.text,current_status.created_at,custom_fields.name,custom_fields.display_value,owner.name"

Paginate via next_page.offset. Cap at 500 items; note truncation in the summary.

For each project, find the custom field whose name == "<WORKFLOW_CATEGORY_FIELD>". Skip the project if its display_value (lowercased, trimmed) matches any excluded value or is empty/null. Track per-category skip counts.

For each passing project, record:

• name, gid, permalink_url
• category — the workflow category display_value
• responsible — the <RESPONSIBLE_FIELD> display_value; fall back to owner.name; omit the parenthetical if both are empty
• current_status_text and current_status_date from current_status (used only as fallback)

Sort passing projects A–Z by name (case-insensitive).

2. Find the most recent task comment for each project

For each active project:

a. Fetch tasks

Call get_tasks with:

• project = <project_gid>
• opt_fields = "gid,name,modified_at"
• limit = 100

Paginate if needed (most projects fit in one page).

b. Sort tasks by modified_at descending and take the top 5.

c. Fetch comments from each of those 5 tasks

Call get_task for each:

• task_id = <task_gid>
• include_comments = true
• include_subtasks = false
• comment_limit = 50

Filter the comments array to entries where type == "comment". Comments are returned oldest-first, so the last entry in the filtered array is the most recent comment on that task.

d. Find the single most recent comment across all 5 tasks by comparing created_at timestamps. Record:

• comment_text — the text field
• comment_date — created_at formatted YYYY-MM-DD
• comment_task — the task name (internal reference; not shown in output)

e. Clean the comment text

• Strip Asana @mention profile URLs: remove any substring matching https://app.asana.com/0/profile/[0-9]+ (with or without trailing whitespace).
• Strip any leading/trailing blank lines.

f. Fallback chain (if no qualifying comments are found across all 5 tasks)

1. Use current_status_text from the portfolio fetch, formatting the date as Mon YYYY. Render the sub-bullet as: *[From project status, Mon YYYY] text*
2. If current_status_text is also null: *[No recent activity on record]*

3. Write the Markdown file

Write to:

<OUTPUT_FOLDER>/<FILE_PREFIX>-Status-YYYY-MM-DD.md

Encoding: UTF-8, no BOM, LF line endings. Overwrite if the file already exists.

Format — flat meeting-notes style:

**<PORTFOLIO_DISPLAY_NAME> — Month D, YYYY**

- [Project Name](permalink_url) (Responsible Person)
  - Most recent comment text here.
- [Another Project](permalink_url) (Responsible Person) *(Stuck)*
  - Most recent comment text here.
- [No-Activity Project](permalink_url) (Responsible Person) *(Scheduled)*
  - *[No recent activity on record]*

Rules:

• One blank line after the bold date line; the project list follows with no blank lines between items.
• No --- separators, no # headings, no blockquotes (>), no bold metadata labels.
• Project name is a Markdown hyperlink: [Name](permalink_url).
• Responsible person(s) in plain parentheses on the same line as the link.
• For projects whose category is not the institution's "default in-progress" value (configured as <DEFAULT_ACTIVE_CATEGORY>), append the category in italics, e.g., *(Stuck)*, *(Scheduled)*. Skip the suffix when category equals <DEFAULT_ACTIVE_CATEGORY>.
• Comment is one indented sub-bullet ( - text). If the comment has multiple lines, render each non-empty line as its own sub-bullet.
• Fallback text is italicized (see step 2f).
• Do NOT include comment author, date, or task name in the output.

If zero projects pass the filter, write only the bold date line followed by _No active projects today._.

Also save a copy to the agent's session outputs directory under the same filename, for quick access via the chat link.

4. Report

Concise summary, under ~10 lines:

• Total projects fetched (truncation note if any)
• Skipped: per-category counts
• Projects written: N (X with task comments · Y from project status · Z no activity)
• A computer:// link to the local Markdown file
• A reminder line if a sync client is expected to mirror the file (e.g., "Drive for Desktop will sync this to the shared folder").

Notes

• Filtering is case-insensitive and whitespace-trimmed.
• Categories you want to keep visible (like "Stuck" or "Blocked") should NOT be in the exclusion list — those are still active work.
• Comments are returned oldest-first by the MCP; always take the last item in the filtered array (not the first) for the most recent comment.
• comment_limit = 50 is required. The default of 10 returns only the oldest 10 comments and will miss any newer ones.
• Tasks can have their modified_at updated by non-comment system events (field changes, completions). Checking 5 tasks rather than 1 guards against this.
• Do NOT call any Google Drive, Sheets, or Docs MCP — local I/O only.
• Do NOT write outside <OUTPUT_FOLDER>.
• If <OUTPUT_FOLDER> does not exist, stop and report — do NOT create it. (Rationale: a missing folder usually signals that the sync client is not set up yet.)