AgentSkillsCN

ceos-ids

在使用EOS IDS流程识别、讨论或解决问题时使用此功能。

SKILL.md
--- frontmatter
name: ceos-ids
description: Use when identifying, discussing, or solving issues using the EOS IDS process

ceos-ids

Run the IDS (Identify, Discuss, Solve) process — the EOS method for resolving issues systematically. Every issue gets a root cause, a focused discussion, and at least one To-Do with an owner and due date.

When to Use

  • •"We have an issue" or "add to the issues list"
  • •"IDS this" or "let's identify discuss solve"
  • •"What are our open issues?" or "show the issues list"
  • •"Solve [issue]" or "discuss [issue]"
  • •"Close this issue" or "mark [issue] as solved"
  • •During L10 meetings when the IDS section begins
  • •Any time the team needs structured problem-solving

Context

Finding the CEOS Repository

Search upward from the current directory for the .ceos marker file. This file marks the root of the CEOS repository.

If .ceos is not found, stop and tell the user: "Not in a CEOS repository. Clone your CEOS repo and run setup.sh first."

Key Files

FilePurpose
data/issues/open/Open issues awaiting resolution
data/issues/solved/Resolved issues (archive)
templates/issue.mdTemplate for new issue files

Issue File Format

Each issue is a markdown file with YAML frontmatter:

yaml
id: issue-001
title: "Slow customer onboarding"
priority: 1            # 1 (highest) - 3 (lowest)
category: process      # people | process | data | vision | traction
ids_stage: identified  # identified | discussed | solved
created: "2026-02-06"

Priority levels:

  • •1 — Critical, solve first
  • •2 — Important, solve soon
  • •3 — Minor, solve when time allows

Categories (the 5 EOS issue types):

  • •people — right person, right seat issues
  • •process — systems, workflows, efficiency
  • •data — metrics, reporting, information gaps
  • •vision — alignment with V/TO, Core Focus
  • •traction — execution, accountability, follow-through

IDS stages:

  • •identified — root cause found, not yet discussed
  • •discussed — perspectives captured, not yet solved
  • •solved — To-Dos created, file moved to data/issues/solved/

File naming: issue-NNN-slug.md where NNN is a zero-padded ID and slug is the title slugified.

Process

Mode: List Issues

When the user asks to see the issues list.

  1. •Read all files in data/issues/open/.
  2. •Parse frontmatter for priority, category, and IDS stage.
  3. •Display sorted by priority (1 first):
code
Open Issues (5 total):
━━━━━━━━━━━━━━━━━━━━
P1  issue-003  Slow customer onboarding     [process]  identified
P1  issue-007  Key account churning          [people]   discussed
P2  issue-001  Reporting gaps in dashboard   [data]     identified
P2  issue-005  Misaligned marketing spend    [vision]   identified
P3  issue-002  Office Wi-Fi unreliable       [process]  identified

If no open issues exist: "No open issues. Nice work! Use 'ceos-ids' to create one when a new issue surfaces."

Mode: Create Issue (Identify)

Use when a new issue surfaces.

Step 1: Name the Issue

Ask: "What's the issue? Give it a short title."

Step 2: Dig for Root Cause (5 Whys)

Don't accept the first answer at face value. The stated problem is usually a symptom.

Walk through the 5 Whys:

  1. •"Why is this happening?" → [answer]
  2. •"And why is that?" → [answer]
  3. •"What's causing that?" → [answer]
  4. •Continue until you reach a root cause (usually 3-5 rounds)

Record both the stated problem and the root cause.

Step 3: Classify

  1. •Priority — How urgent? (1/2/3)
  2. •Category — Which EOS type? (people/process/data/vision/traction)

If the user isn't sure about category, help with examples:

  • •"Is this about having the right people?" → people
  • •"Is this about how work gets done?" → process
  • •"Is this about missing information?" → data
  • •"Is this about our direction?" → vision
  • •"Is this about executing the plan?" → traction

Step 4: Generate the ID

Read existing issue files in data/issues/open/ and data/issues/solved/. Find the highest issue-NNN ID across both directories and increment. If no files exist, start at issue-001.

Step 5: Write the File

Use templates/issue.md as the template. Fill in:

  • •Frontmatter: id, title, priority, category, ids_stage=identified, created=today
  • •Identify section: stated problem and root cause from 5 Whys

Write to data/issues/open/issue-NNN-slug.md.

Show the file before writing. Ask: "Create this issue?"

Mode: Discuss

Use when an identified issue needs discussion (often during L10).

Step 1: Open the Issue

Read the issue file. Show the current Identify section.

Step 2: Capture Perspectives

For each participant:

  • •"What's your perspective on this?"
  • •Record each person's view in the Discuss section

Rules for discussion:

  • •Everyone speaks. No one dominates.
  • •Stay focused on this issue — no tangents.
  • •No solving yet — just understanding.

Step 3: Update the File

  1. •Update ids_stage to discussed in frontmatter
  2. •Fill in the Discuss section with perspectives
  3. •Show the diff, ask for approval, write

Mode: Solve

Use when a discussed issue is ready for resolution.

Step 1: Open the Issue

Read the issue file. Show the Identify and Discuss sections for context.

Step 2: Determine the Solution

Ask: "What's the solve? What specific actions will resolve this?"

For each action:

  1. •Action — what needs to be done (specific and verifiable)
  2. •Owner — one person (not "the team")
  3. •Due date — when will this be done

Step 3: Validate

  • •At least one To-Do is required. An issue can't be "solved" without a concrete action.
  • •Each To-Do has an owner and due date.
  • •Due dates should be realistic (usually within 1-2 weeks).

Step 4: Update the File

  1. •Fill in the Solve section with the To-Do table
  2. •Update ids_stage to solved in frontmatter
  3. •Show the diff, ask for approval, write

Step 5: Move to Solved

After writing the solved issue:

  1. •Move the file from data/issues/open/ to data/issues/solved/
  2. •Confirm: "Issue [title] moved to solved. To-Dos are now the owners' responsibility."

If the user wants to keep it in open until To-Dos are done, that's fine — the ids_stage: solved status still indicates the IDS process is complete.

Mode: Full IDS (Inline)

Use during L10 meetings or when running all three stages back-to-back.

  1. •Identify — Run 5 Whys, classify
  2. •Discuss — Capture perspectives
  3. •Solve — Create To-Dos
  4. •Close — Move to solved

This is the same as the three modes above, just run sequentially on a single issue.

Output Format

List: Priority-sorted table of open issues. Create: Complete issue file with Identify section. Discuss: Updated issue file with perspectives added. Solve: Completed issue file with To-Do table, moved to solved directory.

Guardrails

  • •Always show diff before writing. Never modify an issue file without showing the change and getting approval.
  • •At least one To-Do to solve. An issue without an action item isn't solved — it's just discussed. Push for a concrete next step.
  • •One owner per To-Do. Not "the team" or "everyone." One person is accountable.
  • •Don't skip Identify. The 5 Whys matter. The most common failure is solving the symptom instead of the root cause. Push past the first answer.
  • •Respect the stages. Don't jump to Solve without first Identifying and Discussing. The stages exist to prevent premature solutions.
  • •ID uniqueness. Check both open/ and solved/ directories when generating a new ID to avoid collisions.
  • •File lifecycle. Issues start in data/issues/open/, move to data/issues/solved/ when the IDS process is complete. Don't delete issue files — solved issues are the audit trail.
  • •Integrate with L10. During L10 meetings, IDS is Section 6. The ceos-l10 skill will surface the issues list — this skill handles the actual IDS work on individual issues.