agentleFS
eugenezhangco-speceugenezhangco-spec/athena/.claude/skills/eng-inbox/SKILL.md

eng-inbox

Skill0 starsChanged 7 months agoMITView on GitHub

When this skill is triggered, perform ALL of the following steps: List all files in context/inbox/. If empty, say "Inbox is empty — nothing to process" and stop. For each file in the inbox: - PDFs: Read and extract all content - Word docs (.docx): Read and extract all content - Excel files (.xlsx): Read every sheet, every row, every cell - PowerPoint (.pptx): Read all slides and speaker notes - Email threads (.pdf/.eml): Read full thread, identify key decisions…

The file

Show its rules and text

We split this skill into the rules it gives an agent

22 rules. About 0 are common, 13 are distinctive, and 9 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈13

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 9

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    When this skill is triggered, perform ALL of the following steps:

    line 3
  2. Distinctive:

    PDFs: Read and extract all content

    line 12
  3. Distinctive:

    Word docs (.docx): Read and extract all content

    line 13
  4. Distinctive:

    Excel files (.xlsx): Read every sheet, every row, every cell

    line 14
  5. Distinctive:

    PowerPoint (.pptx): Read all slides and speaker notes

    line 15
  6. Very specific:

    Email threads (.pdf/.eml): Read full thread, identify key decisions and action items

    line 16
  7. Distinctive:

    Markdown/text: Read in full

    line 17
  8. Distinctive:

    Keep the briefing concise — just the highlights.

    line 28
  9. Distinctive:

    If any inbox items contain new requirements, scope changes, feature requests, architectural decisions, or anything that affects what gets …

    line 32
  10. Very specific:

    Write a structured handoff file to context/sessions/YYYY-MM-DD-HHMM-inbox-[topic].md

    line 34
  11. Very specific:

    Use the same format as a /wrap handoff (same frontmatter schema)

    line 35
  12. Very specific:

    Create new subfolders: if needed. The structure is dynamic — if a file doesn't fit existing folders, create an appropriate one (e.g., …

    line 68
  13. Very specific:

    Read context/STATUS.md and update with any new information:

    line 72
  14. Distinctive:

    If any inbox files contain technical specifications, schema changes, API details, or architecture decisions:

    line 80
  15. Very specific:

    Update context/ENGINEERING.md with the new technical details

    line 81
  16. Distinctive:

    If new context changes the project direction, adds new concepts, or modifies how things work:

    line 86
  17. Very specific:

    Update context/EXPLAINED.md so the plain-English reference stays current

    line 87
  18. Distinctive:

    Keep the plain-English style and teaching tone

    line 88
  19. Very specific:

    Read context/ACTION-ITEMS.md and add any new action items discovered during inbox processing:

    line 92
  20. Distinctive:

    Follow-ups from communications ("get back to [person] about X")

    line 95
  21. Very specific:

    Create a markdown file in context/communications/ with the date and topic

    line 134
  22. Distinctive:

    Update STATUS.md and memory as needed

    line 135

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

/inbox — Process New Context

When this skill is triggered, perform ALL of the following steps:

Step 1: Check the inbox

List all files in context/inbox/. If empty, say "Inbox is empty — nothing to process" and stop.

Step 2: Read every file

For each file in the inbox:

  • PDFs: Read and extract all content
  • Word docs (.docx): Read and extract all content
  • Excel files (.xlsx): Read every sheet, every row, every cell
  • PowerPoint (.pptx): Read all slides and speaker notes
  • Email threads (.pdf/.eml): Read full thread, identify key decisions and action items
  • Markdown/text: Read in full
  • Images/screenshots: View and describe what's shown

Step 3: Brief in plain English

For each file, provide:

  • What it is (1 sentence)
  • What's new or important (2-3 bullets)
  • Does it change our current plan? (yes/no + why)
  • Action items (if any)

Keep the briefing concise — just the highlights.

Step 3b: Write AI Build Brief (if build-relevant)

If any inbox items contain new requirements, scope changes, feature requests, architectural decisions, or anything that affects what gets built:

  1. Write a structured handoff file to context/sessions/YYYY-MM-DD-HHMM-inbox-[topic].md
  2. Use the same format as a /wrap handoff (same frontmatter schema)
  3. Set status: ready so /pull can load it

File content must include:

  • What This Session Was About — why this inbox item matters to the build
  • What Was Decided — any decisions or direction changes derived from the content, written as directives
  • Build Brief — what needs to be built or changed, written for AI consumption. Specific and discrete. A fresh session must be able to plan from this without reading the original document.
  • Constraints and Boundaries — what must not change, what assumptions must not be made
  • Open Questions — anything that needs a decision before planning can start
  1. Say: "I've written a build brief to context/sessions/[filename]. Run /pull in your build session to load it into planning."

Do not write a build brief for purely informational items (stakeholder profiles, historical docs, emails with no action items). Only write one when the inbox content has direct build implications.

Step 4: Sort files to proper locations

Move each file from context/inbox/ to the appropriate folder:

File typeDestination
Business plans, historical docscontext/archive/business-plans/
Technical specs, briefscontext/archive/technical-briefs/
Pitch decks, presentationscontext/archive/pitch-decks/
Research, analysis, reportscontext/archive/research/
Prep docs, confidentialcontext/archive/prep-docs/
Media files (audio, video)context/archive/media/
Email threads, meeting notescontext/communications/
Requirements, schemas, strategiescontext/requirements/
Stakeholder profilescontext/stakeholders/

Rename files to be descriptive and kebab-case. Remove UUIDs and cryptic names. Example: 6a485b84-e637-48dc...pdf → stakeholder-call-prep-march-2026.pdf

Create new subfolders if needed. The structure is dynamic — if a file doesn't fit existing folders, create an appropriate one (e.g., context/legal/, context/funding/, context/[partner-name]/).

Step 5: Update STATUS.md

Read context/STATUS.md and update with any new information:

  • New decisions or direction changes
  • Updated timelines or milestones
  • New stakeholders or relationships
  • Changed priorities or scope

Step 6: Update ENGINEERING.md (if technical content)

If any inbox files contain technical specifications, schema changes, API details, or architecture decisions:

  • Update context/ENGINEERING.md with the new technical details
  • Note what changed and why in a changelog entry

Step 7: Update EXPLAINED.md (if it affects the owner's understanding)

If new context changes the project direction, adds new concepts, or modifies how things work:

  • Update context/EXPLAINED.md so the plain-English reference stays current
  • Keep the plain-English style and teaching tone

Step 8: Update ACTION-ITEMS.md

Read context/ACTION-ITEMS.md and add any new action items discovered during inbox processing:

  • Action items explicitly mentioned in documents (e.g., "please review", "need to schedule", "follow up on")
  • Implied work from new requirements or scope changes
  • Follow-ups from communications ("get back to [person] about X")
  • Anything flagged in the Step 3 briefing as needing action

For each new item, use the next available ID from the next_id counter, increment the counter, and follow the format defined in ACTION-ITEMS.md. Set the Source to the inbox file name and date.

Also check if any existing action items are resolved by the new context (e.g., an answer arrived for an open question). If so, mark them as done.

Step 9: Update memory

Save any important new context to the memory system:

  • New stakeholder information → update or create stakeholder memory
  • New project decisions → update project memories
  • New technical decisions → update technical decision memories
  • Changed timelines or priorities → update status memories

Step 10: Confirm

After processing, provide:

  • Summary of what was processed
  • Where each file was moved
  • What was updated (STATUS.md, ENGINEERING.md, EXPLAINED.md, ACTION-ITEMS.md, memory)
  • Action items: N new items added, N existing items updated
  • Any questions or decisions that need input

When to trigger

This skill runs when you say any of:

  • "/inbox"
  • "check the inbox"
  • "new stuff"
  • "read the updates"
  • "there's new files"
  • "I dropped something in the inbox"
  • "process the inbox"

Verbal context

Verbal context may also be provided in chat instead of dropping files. When notes from conversations, calls, lunches, or text messages are shared:

  1. Capture the key points
  2. Create a markdown file in context/communications/ with the date and topic
  3. Update STATUS.md and memory as needed
  4. Brief on what was captured to confirm accuracy
  5. If the verbal context has build implications (new requirements, scope changes, feature requests, direction changes): also apply Step 3b — write an AI build brief to context/sessions/YYYY-MM-DD-HHMM-verbal-[topic].md with status: ready. Say: "I've written a build brief to context/sessions/[filename]. Run /pull in your build session to load it into planning."

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.