Skip to content
Back to skills

Team Health

ASecurity

Keeping the team (of agents and people) effective and sustainable: minimal process, retrospectives, unblocking, and continuous improvement. Trigger: load when designing/adjusting process, running a retro, addressing recurring friction, or when the team's way of working needs tuning.

  • 2 stars
  • 0 votes
  • 0 copies
  • 0 views
  • Added October 13, 2026
ai-agentsgoaws

Security analysis

A100/100

Scanned October 13, 2026

npx -y skills add jrodrigopuca/local-agents --skill team-health --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Team Health?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Team Health
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/jrodrigopuca-team-health/badge)](https://www.skillsdirectory.com/skills/jrodrigopuca-team-health)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: team-health
description: >
  Keeping the team (of agents and people) effective and sustainable: minimal
  process, retrospectives, unblocking, and continuous improvement. Trigger:
  load when designing/adjusting process, running a retro, addressing recurring
  friction, or when the team's way of working needs tuning.
license: Apache-2.0
metadata:
  author: jrodrigopuca
  version: "1.1"
---

## Critical Patterns

### 1. Process is a tool, not a religion — the minimum that helps

Every ritual, meeting, and template must earn its place by making the work
better; the moment it's followed for its own sake, cut it. Process exists to
serve delivery and clarity, not to perform organization. The right amount is the
least that keeps the work coordinated and visible — a step past that is pure tax
([orchestration's](../orchestration/SKILL.md) "know when NOT to orchestrate",
at the process level).

### 2. Fix the system, not the person — recurring problems are design flaws

When the same mistake keeps happening, the cause is almost never carelessness —
it's a missing guardrail, an unclear interface, or a process gap. Same blameless
DNA as [stark/crisis-mode](../../../stark/skills/crisis-mode/SKILL.md) and
[qa](../../../qa/AGENTS.md): "someone forgot" is not a root cause; "there was no
check that would have caught it" is. Ask what SYSTEM change makes the mistake
hard to repeat — a template, an automated check, a clearer handoff — rather than
asking people to try harder.

### 3. Retrospectives produce changes, not venting

A retro that ends without a small number of owned, concrete changes was group
therapy. The shape: what went well (keep doing), what hurt (with the real
cause, not the symptom), and 1-3 changes with an owner and a next-cycle check.
Few changes actually adopted beat twenty listed and forgotten — the
[architect's mentoring](../../../architect/skills/mentoring/SKILL.md) "one
lesson per moment" applied to team learning.

### 4. Sustainable pace — crunch borrows from next week at high interest

Sustained overload doesn't deliver more; it delivers more BUGS and burnout,
which cost more than they saved. Crisis-mode intensity is for actual crises
([borrowed authority that gets repaid (`crisis-mode`)](../../../stark/skills/crisis-mode/SKILL.md)),
not the default operating mode. A team (or an agent pipeline) run permanently
hot has a planning problem wearing an effort costume — the fix is upstream, in
scope and WIP, not in asking for more hours.

### 5. A roster of agents has its own recurring frictions — and they are measurable

The team here is agents as much as people, and their friction doesn't show up
as a bad mood. It shows up as: a handoff nobody received (the skill on the
other side has nothing to say), an agent that stops loading its own skill and
answers from the web instead, a promise that holds three times out of four (a
flaky scenario is an inconsistent agent or a criterion measuring the host, never
"try again"), verification skipped when the run got long, a specialist quietly
substituted for one the user named. Every one of those is #2 in disguise — a
missing check, not a careless agent — and the fix is a scenario or a validator
rule, not a reminder in a prompt. Feedback to people stays specific and close
to the event (what happened, its effect, what to change), and flows UP too: the
manager who can't hear "this process slows us down" is optimizing blind.

### 6. Improve in increments a scenario can confirm

Team health is a habit, not a project: notice one friction, file off one edge,
confirm it stayed fixed. For the agents that means one change ships with the
check that would catch its regression; for the process it means one clearer
handoff or one deleted ritual per cycle, never a reorganization — a reorg is a
big-bang deploy with people in it, high-risk and slow to roll back. Small fixes
that stay fixed outrun any heroic overhaul, and "stayed fixed" is the part you
measure.

## Resources

- Sibling skills: [delivery](../delivery/SKILL.md) (sustainable pace makes
  delivery predictable), [orchestration](../orchestration/SKILL.md) (minimal
  coordination)
- Blameless/systemic thinking shared across the team:
  [stark/crisis-mode](../../../stark/skills/crisis-mode/SKILL.md),
  [qa](../../../qa/AGENTS.md)

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.