psychology-agent: Command for Claude Code

/capacity — Cognitive Architecture Capacity Assessment

Assess remaining capacity across every constrained dimension of the cognitive architecture. Run at session start, when approaching a new phase of work, or when the system feels crowded.

---

Dimensions to Assess

1. MEMORY.md (hard constraint)

PROJECT_ROOT="$(git rev-parse --show-toplevel 2>/dev/null || pwd)"
_HASH="$(echo "$PROJECT_ROOT" | tr '/' '-')"
wc -l "$HOME/.claude/projects/${_HASH}/memory/MEMORY.md"

Threshold	Status	Action
< 185	Healthy	No action needed
185–199	Pressure	Move stable content to CLAUDE.md before adding
200	At limit	Mandatory archival before any new content
> 200	OVERFLOWING	System is silently truncating — fix immediately

Report: N / 200 lines (N remaining before pressure / hard limit)

2. CLAUDE.md (advisory constraint)

PROJECT_ROOT="$(git rev-parse --show-toplevel 2>/dev/null || pwd)"
wc -l "${PROJECT_ROOT}/CLAUDE.md"

Advisory limit ~200 lines. Holds stable conventions — grows slowly. Report: N lines used (~M available before advisory limit)

3. cognitive-triggers.md (practical constraint)

PROJECT_ROOT="$(git rev-parse --show-toplevel 2>/dev/null || pwd)"
wc -l "${PROJECT_ROOT}/docs/cognitive-triggers.md"

Loaded in full at T1. Practical ceiling ~900–1000 lines before context loading becomes a meaningful tax. Above 800: flag for architectural review (consider splitting into core + extended). Report: N lines (~M before practical ceiling)

4. Trigger Coverage

List current triggers T1–TN and identify gaps:

T1  — Session start
T2  — Before any response
T3  — Before any recommendation
T4  — Before writing to disk
T5  — Gap check (phase boundaries)
T6  — After user pushback
T7  — After user approval
T8  — After task completion
T9  — Memory hygiene
T10 — Lesson surfaces
T11 — Cogarch self-audit
T12 — Positive pattern recognition
T13 — External content entering context