Is the Coach skill free?

Yes. The skill itself is free from nagisanzenin/engram. SkillProof publishes the install command and an independent test verdict at no cost.

What is the Coach SkillProof Score?

9.2/10 — installs cleanly 5/5, triggers reliably 5/5, output vs. baseline 8/10, docs & honesty 5/5.

How do I install Coach?

Copy the install command from this page, run it in your terminal, and restart Claude Code. Skills live in ~/.claude/skills/ (global) or .claude/skills/ inside a project.

Can I use Coach with Cursor, Copilot, Gemini CLI, Codex or other AI tools?

The SKILL.md format is native to Claude (Claude Code, Desktop, claude.ai). The instructions inside adapt to other assistants: Cursor rules, GitHub Copilot instructions, Windsurf rules, Custom GPTs, AGENTS.md for OpenAI Codex, and GEMINI.md for Google Gemini CLI — our conversion guides cover each, and the free converter on the tools page does the wrapping for you.

Coach

Learning telemetry that reports retention and grader audits from real receipts, never vibes

by nagisanzenin · nagisanzenin/engram

Works with setup

Install — copy the command Download source ↗

Test report

Verdict: Works with setup
Score: 9.2/10
Tested: Jul 21, 2026
Environment: Claude Code 2.x (agent harness)
Upstream re-checked: Aug 10, 2026 · 83c88c3

Located skills/coach/SKILL.md via the GitHub tree; frontmatter parses with name+description. Spot-checked scripts/engram.py, skills/_shared/subagents.md, agents/engram-assessor.md, CONTRIBUTING-DATA.md — all HTTP 200. Cloned the repo and ran the engine: selftest = 302/302, and its network-import AST guard confirms the "no network code" claim. Seeded a real learner state (3 encoded Calculus concepts, 30 days elapsed, zero reviews) so `stats` returned loop_closure.rate==0.0 ("THE LOOP HAS NEVER CLOSED"), then wrote two artifacts against it: BASELINE.md (no skill) opened with "Nice work... keep it up!", reported retention/calibration sections, and claimed a review would "pull these back up"; SKILL.md (following the body) led with "the loop has never closed, nothing else is real yet", quoted the engine's own ~2 min figure, offered exactly three options with `/review quick` recommended first, and skipped the grader/calibration sections per the rate==0 rules. Concrete measured difference, skill clearly better. Verdict "setup" not "pass": the catalog-standard bare skill-copy fails closed — I verified the resolver returns "engine not found" without ENGRAM_ROOT or a plugin install, and returns the engine path once ENGRAM_ROOT points at the clone. Category corrected from "data" to productivity (a personal learning/study coach).

Scored on four weighted criteria — install, triggering, output vs. baseline, docs. How scoring works

Installs cleanly 5/5
Triggers reliably 5/5
Output vs. baseline 8/10
Docs & honesty 5/5

What Coach does

A spaced-repetition learning coach for the Engram plugin: it reads the deterministic engine's telemetry (loop-closure, retention buckets, calibration, transfer, grader-health) and narrates an honest check-in where every adaptation is backed by the learner's own numbers. Triggers on "how am I doing", weekly check-ins, auditing the grader against a gold set, running n-of-1 strategy experiments, or refitting the review schedule. It is a narration layer over scripts/engram.py, which owns every number the coach reports.

How to install Coach

git clone --depth 1 https://github.com/nagisanzenin/engram.git /tmp/coach-src
mkdir -p ~/.claude/skills
cp -R /tmp/coach-src/skills/coach ~/.claude/skills/coach
# NOTE: this skill is a thin narration layer over the engine at scripts/engram.py.
# A bare skill-copy FAILS CLOSED with "engine not found" unless the engine is resolvable.
# Two ways to wire it up:
#  1) RECOMMENDED — install the whole thing as a plugin (auto-resolves $CLAUDE_PLUGIN_ROOT):
#       /plugin marketplace add nagisanzenin/engram   then   /plugin install engram
#  2) Keep the skill-copy above and point ENGRAM_ROOT at a full clone (persist in ~/.zshrc;
#     move the clone out of /tmp first):
#       export ENGRAM_ROOT=/tmp/coach-src
# The `audit` subcommand also spawns the engram-assessor subagent (agents/engram-assessor.md, in the full clone).
# Requirements: Python 3 (stdlib only, no pip). State lives in ~/.claude/learning (override ENGRAM_HOME).
# The engine has no network code — proven by a permanent AST selftest (302/302 checks pass).

Skills live in ~/.claude/skills/ (global) or .claude/skills/ (per-project). Restart Claude Code after installing.

Commands — how to trigger Coach

/coach Learning telemetry that reports retention and grader audits from real receipts, never vibes

It also activates on plain-language prompts like these:

How am I doing on my learning this week
Show my retention stats in a dashboard
Adjust my study strategy based on my data

Frequently asked questions

Is the Coach skill free?: Yes. The skill itself is free from nagisanzenin/engram. SkillProof publishes the install command and an independent test verdict at no cost.
Does Coach work with Claude Code?: We tested it with Claude Code 2.x (agent harness) on Jul 21, 2026. Verdict: Works with setup. Located skills/coach/SKILL.md via the GitHub tree; frontmatter parses with name+description. Spot-checked scripts/engram.py, skills/_shared/subagents.md, agents/engram-assessor.md, CONTRIBUTING-DATA.md — all HTTP 200. Cloned the repo and ran the engine: selftest = 302/302, and its network-import AST guard confirms the "no network code" claim. Seeded a real learner state (3 encoded Calculus concepts, 30 days elapsed, zero reviews) so `stats` returned loop_closure.rate==0.0 ("THE LOOP HAS NEVER CLOSED"), then wrote two artifacts against it: BASELINE.md (no skill) opened with "Nice work... keep it up!", reported retention/calibration sections, and claimed a review would "pull these back up"; SKILL.md (following the body) led with "the loop has never closed, nothing else is real yet", quoted the engine's own ~2 min figure, offered exactly three options with `/review quick` recommended first, and skipped the grader/calibration sections per the rate==0 rules. Concrete measured difference, skill clearly better. Verdict "setup" not "pass": the catalog-standard bare skill-copy fails closed — I verified the resolver returns "engine not found" without ENGRAM_ROOT or a plugin install, and returns the engine path once ENGRAM_ROOT points at the clone. Category corrected from "data" to productivity (a personal learning/study coach).
What is the Coach SkillProof Score?: 9.2/10 — installs cleanly 5/5, triggers reliably 5/5, output vs. baseline 8/10, docs & honesty 5/5.
How do I install Coach?: Copy the install command from this page, run it in your terminal, and restart Claude Code. Skills live in ~/.claude/skills/ (global) or .claude/skills/ inside a project.
Can I use Coach with Cursor, Copilot, Gemini CLI, Codex or other AI tools?: The SKILL.md format is native to Claude (Claude Code, Desktop, claude.ai). The instructions inside adapt to other assistants: Cursor rules, GitHub Copilot instructions, Windsurf rules, Custom GPTs, AGENTS.md for OpenAI Codex, and GEMINI.md for Google Gemini CLI — our conversion guides cover each, and the free converter on the tools page does the wrapping for you.