원클릭으로
aim-parzival-loader
Two-phase capped startup context loaders for Parzival (activation + session-start)
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
메뉴
Two-phase capped startup context loaders for Parzival (activation + session-start)
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
SOC 직업 분류 기준
| name | aim-parzival-loader |
| description | Two-phase capped startup context loaders for Parzival (activation + session-start) |
| allowed-tools | Bash |
Two thin, deterministic loader scripts that emit Parzival's startup context as a
single consolidated, capped block per phase — replacing the scattered
per-file reads at activation and session-start. Caps follow the approved A2 cap
table (oversight/tasks/task077-startup-governance/A2-CAP-TABLE-APPROVED.md).
The full files always remain on disk at full size; the loaders emit capped
heads / recency slices / pointers, never destructive edits.
Caps are enforced by line / byte / entry counts only — no tokenizer at
runtime. Per-phase token budgets are asserted separately by
tests/unit/test_parzival_loaders.py.
/pov:parzival, Tier A)Run once during activation (after the sanctum presence-check / self-heal):
python3 "${AI_MEMORY_INSTALL_DIR:-$HOME/.ai-memory}/_ai-memory/pov/skills/aim-parzival-loader/activation_loader.py" "$(pwd)"
Emits, in the approved A2 order: config.yaml (full) → project-status.md
(head-capped to Contract 60 lines / 6 KB) → global constraints (full) → phase
constraints (full; phase from project-status) → CREED.md (full) →
PERSONA.md (## Evolution Log capped to the last 10 identity-shift rows) →
BOND.md First-Breath marker-scan only (not a load) → WORKFLOW-MAP.md
(full).
/pov:parzival-start, Tier B + oversight)# oversight summaries (step-01)
python3 ".../aim-parzival-loader/session_loader.py" "$(pwd)" --scope oversight
# sanctum Tier B, after the bootstrap Qdrant section (step-01b section 4)
python3 ".../aim-parzival-loader/session_loader.py" "$(pwd)" --scope sanctum
--scope honors the A2 interleave (oversight → bootstrap Qdrant → sanctum):
oversight — SESSION_WORK_INDEX (full) + tracking active sections + bugs/TD
## Quick Stats only.sanctum — LORE.md recency-weighted slice (structural sections + newest
"Things Learned" lessons ≤ 25 KB + pointer) → BOND.md full (vital floor) →
sanctum MEMORY.md full.all — oversight then sanctum (used by the budget / live-smoke tests).This loader never reads a SESSION_HANDOFF_*.md file — the most recent handoff
comes from bootstrap L1; the file is read only on the existing CASE-A/CASE-B
fallback (step-01b).
activation_loader.py, session_loader.py, shared loader_common.pyMEMORY.md files — do not conflate: the sanctum
_ai-memory/sanctum/parzival/MEMORY.md (Tier-B identity, loaded here) is NOT
the Claude-Code auto-memory ~/.claude/projects/<slug>/memory/MEMORY.md.(absent) markers); startup
never blocks on a missing sanctum/oversight file.Detect content drift of an operator's scaffolded sanctum files (BOND, CAPABILITIES, CREED, INDEX, LORE, MEMORY, PERSONA, PULSE) against the evolving reference templates, and surface recommended add/remove WITH rationale — never a silent overwrite. Use on a session-start drift check, after the reference templates change, or when the operator asks whether their sanctum is current.
Check ai-memory system status and collection stats
Check ai-memory system status and collection stats
Manually save current session context to ai-memory
Search ai-memory for relevant stored memories
Search ai-memory for relevant stored memories