cd /news/ai-tools/show-hn-an-ai-coaching-team-in-claud… Β· home β€Ί topics β€Ί ai-tools β€Ί article
[ARTICLE Β· art-25611] src=github.com β†— pub= topic=ai-tools verified=true sentiment=Β· neutral

Show HN: An AI coaching team in Claude Code that's forbidden from judging you

A developer released Intelligence Emotions, an open-source AI coaching team that runs inside Claude Code to help users build mental fitness by identifying self-sabotaging thought patterns and practicing attention shifts. The tool, which is not a substitute for therapy, uses five AI coaches and a private local journal to track patterns over time.

read6 min publishedJun 12, 2026

Your mind is your best friend β€” and, left unsupervised, your worst critic.

Intelligence Emotions turns Claude Code into your personal mental-fitness team, built on the Intelligence Emotions model. Five coaches, one practice: catch the ten saboteurs (led by the Judge β€” the inner voice that finds fault with everything), do PQ reps (10-second shifts of attention to a physical sensation), and let the Sage choose the response. The team remembers between sessions β€” a private, local-only journal that makes every coach smarter about your actual patterns.

This is not therapy or medical care. The coaches say so themselves, once, plainly β€” and they know when to stop coaching and point you to a professional instead.

Who this is for:

Anyone building mental fitness who wants the practice to actually happen dailyFirst-time practitionersβ€” structured coaching instead of a blank journal** People with a challenge they keep circling**β€” a habit, a conflict, a decision that won't resolve

  • Install Intelligence Emotions (30 seconds β€” see below)
  • Run /sage-session

β€” bring one real challenge - Run /intercept

the next time a thought stings - Run /pq-retro

tonight before closing the laptop - Run /pq-score

after three days of journal data - Stop there. You'll know if this is for you.

Requirements: Claude Code, Git, Bun v1.0+

Open Claude Code and paste this. Claude does the rest.

Install Intelligence Emotions: run

then add an "Intelligence Emotions" section to CLAUDE.md that lists the available skills: /pq, /sage-session, /intercept, /saboteur-scan, /sage-perspective, /navigate-review, /pq-retro, /daily-pipeline, /growth-spec, /commit, /pq-score, /habit-watch, /life-design, /saboteur-audit, /insight-doc, /second-coach, /context-save, /context-restore β€” and says that /pq is the front door when unsure which to use.git clone --single-branch --depth 1 https://github.com/YOUR-GITHUB-USERNAME/intelligence-emotions.git ~/.claude/skills/pq && cd ~/.claude/skills/pq && ./setup

(The skills use the pq

prefix β€” your mental-fitness quotient, the percentage of time your mind serves you rather than sabotages you.)

/sage-session

Bring something real β€” a challenge with weight, however half-formed. The Sage takes it from there.

You:  /intercept my coworker got promoted and I feel worthless

Spotter:  That word "worthless" β€” let's trace where it came from.
          What's the actual sentence running in your head, word for word?

You:  "she's two years younger and already ahead of me"

Spotter:  Two patterns in that sentence. The comparison clock β€” that's the
          Hyper-Achiever (the pattern that pegs your worth to external
          scorekeeping). And "worthless" is the Judge delivering the verdict.
          The lie: her promotion is a measurement of you. It feels like data.
          It's machinery β€” the same machinery would have produced the same
          sentence about whoever got promoted.

          Next time a promotion email lands β€” anyone's β€” ten seconds on your
          feet on the floor before your mind gets the microphone.
          Want me to log this interception?

The daily loop, as a team:

Time Command Coach What happens
Morning /daily-pipeline
The Trainer One intention, a saboteur forecast for the day's terrain, first rep
In the moment /intercept
The Spotter Trace a stinging thought: trigger β†’ saboteur β†’ the lie β†’ the need. Rep prescribed, interception logged
Evening /pq-retro
The Witness Five minutes: what triggered, what got caught, rep count, one Sage win. Writes the journal
Anytime /sage-session
The Sage The flagship β€” any life challenge, walked through all five Sage powers

The 21-day arc, when something deserves real practice:

Step Command What happens
1 /growth-spec
A vague intention ("be more patient with my kids") becomes a concrete 21-day practice: trigger inventory, daily reps anchored to real moments, check-ins
2 /habit-watch
Scheduled look-ins; catches drift while the fix is still one sentence
3 /commit
Day 21: the honest review, the closing entry, the commitment marked. No grades β€” what happened, what it taught, what's next
Situation Command
A whole situation with weight /sage-session
One specific moment that stung /intercept
"Check this draft/entry before I send it" β€” no fixes wanted /saboteur-scan
A decision I keep flip-flopping on /sage-perspective
"Will this plan actually stick?" /navigate-review
"Map all my patterns" β€” the wide view /saboteur-audit
"How am I doing?" β€” numbers from real data /pq-score
Design my mornings / environment from scratch /life-design
Something just clicked β€” keep it /insight-doc
"Push back on me harder" /second-coach
Not sure /pq

Your data, on the command line (all local, all yours):

~/.claude/skills/pq/bin/pq-journal-search --days 7           # this week's interceptions, entries, commitments
~/.claude/skills/pq/bin/pq-journal-search --stats --days 7   # per-day reps, interception rate, saboteur tallies
~/.claude/skills/pq/bin/pq-journal-log saboteurs '{...}'     # log an interception by hand
~/.claude/skills/pq/bin/pq-journal-log entries --redact <id> # forget a record β€” expunged completely, archive included
~/.claude/skills/pq/bin/pq-config list                       # your preferences (name, rep_target, cadence, explain_level)

The journal is event-sourced and append-only: three streams under ~/.pq/journal/

(entries.jsonl

, saboteurs.jsonl

, commitments.jsonl

), supersede-don't-delete, redact-on-demand. A write-time guard refuses credential-shaped secrets. /pq-score

and /saboteur-audit

read this data β€” nothing is ever estimated.

~/.claude/skills/pq/setup --uninstall

Removes the skill symlinks. Your journal and config in ~/.pq

are never touched β€” that data is yours, including after you leave.

rm -rf ~/.claude/skills/pq
for s in pq sage-session intercept saboteur-scan sage-perspective navigate-review \
         pq-retro daily-pipeline growth-spec commit pq-score habit-watch life-design \
         saboteur-audit insight-doc second-coach context-save context-restore; do
  [ -L ~/.claude/skills/$s ] && rm ~/.claude/skills/$s
done

CLAUDE.mdβ€” contributor guide: the pipeline, the iron rules, how to write a skillETHOS.mdβ€” the operating philosophy: label don't fight; ten seconds at a time; every setback is a gift; the coach is never the JudgeCHANGELOG.mdβ€” versions, starting fresh at 1.0.0.0

There is no telemetry. None, not opt-out β€” none exists. Everything personal lives in ~/.pq/

with owner-only permissions (0700 directories, 0600 files), never pushed to any remote, never sent to any service. Redaction is honored absolutely: ask any coach to forget something and the record is expunged from the live log and the archive. The test suite enforces that no external sink can quietly appear.

Skills don't show up in Claude Codeβ€” re-run~/.claude/skills/pq/setup

; check that~/.claude/skills/sage-session

is a symlink pointing into the install.β€” installbun: command not found

Bun, then re-run./setup

.A coach sounds like the Judgeβ€” that's a bug in the product, not a feature of honesty. Open an issue with the transcript line; the anti-Judge rule is the product.

Exactly 10 saboteurs. Exactly 5 Sage powers. Never an invented eleventh or sixth β€” the test suite enforces the canon. The team: the Sage (lead coach), the Spotter (pattern recognition), the Trainer (rep drills), the Navigator (plans and direction), the Witness (reflection). One rule outranks everything they say: no coach is ever the Judge.

MIT β€” see LICENSE.

── more in #ai-tools 4 stories Β· sorted by recency
sponsored brought to you by zahid.host 4,200+ EU-deployed projects
reading about agents? ship yours in a single git push.

Run your AI side-project on zahid.host

EU-based hosting, git-push deploys, automatic HTTPS, no cold starts. Free tier with a custom domain β€” perfect for shipping the agent you just read about.

$git push zahid main
β†’ Live at https://your-agent.zahid.host βœ“
Get free account β†’ Pricing
from €0/mo Β· no card required
LIVE [news/show-hn-an-ai-coachi…] indexed:0 read:6min 2026-06-12 Β· β€”