Operate an explicitly activated LoopX Material Lifecycle for a connected project. Use for material-store inventory, lossless migration, candidate/archive transitions, exact-read-backed ranking, ranked-entry rebuilds, bounded Explore intake, owner-gated apply, rollback, and audit. Do not use for ordinary one-off reading or research when the project has not activated Material Lifecycle.
88%
Does it follow best practices?
Run evals on this skill
Adds up to 20 points to the overall score
Passed
No findings from the security scan
Use this skill for the lifecycle and authority of a project's durable material store. Source discovery, domain-specific scoring, and note writing may be provided by project skills; this skill owns the generic lossless lifecycle.
LoopX ships the canonical source for this skill, but does not install it into the user's global skill directory. Install a managed copy only in a connected project that explicitly enables Material Lifecycle:
loopx project-skill install \
--project . \
--skill loopx-material \
--surface codex \
--executeUse --surface antigravity, --surface claude-code, --surface kiro-cli,
--surface opencode, or --surface pi for those hosts; repeat the flag to
install multiple host-native copies in one transaction. Managed copies live
under .agents/skills/, .agent/skills/, .claude/skills/, .kiro/skills/,
.opencode/skills/, or .pi/skills/ and are upgraded or removed through the
same CLI. Project-local discovery does not itself activate material-store
writes; the selected execution owner still needs explicit Material Lifecycle
source authority.
Before changing a material store:
Codex,
Antigravity, Claude Code, OpenCode, Kiro CLI, etc.) or an authorized
native Bot turn under the same Core workspace write grant; never run
start-goal, invent a dummy Goal merely to obtain material metadata, or
borrow another host's turn or session identity.loopx project-skill status --project . --skill loopx-material and
confirm the required host surfaces are current.material_lifecycle/the Goal declares its source profile, or the project
source owner declares an active material profile and an existing Core
workspace write grant. A catalog entry, scope reference or skill copy is
not activation.loopx material-lifecycle architecture --format json and preserve its
default-off, owner-gated, provider-neutral boundaries. The project source
route supports inventory, candidate intake/rollback, rerank packets,
readable projection and settlement; it does not widen Goal-only migration,
rebuild or Explore APIs.If the project-local skill is missing, preview an explicit project install; do not fall back to a global copy. If activation or authority is missing, stop before source mutation. Create a bounded setup todo or owner gate; do not invent a store or treat chat history as authority.
Keep these responsibilities separate:
Read the source authority before proposing structural change.
No migration, rebuild, or rerank may begin from an unverified partial parse.
Use stable material references across:
candidate -> active -> archived
\-> carryover
archived -> activeEvery transition needs a revisioned evidence or decision reference. Archiving must preserve the original source reference and an archive reference. Reading, summarizing, or publishing a note does not implicitly archive a material.
Recall is advisory. Before a material affects ranking or lifecycle:
Do not start Explore merely because the current list feels incomplete. Explore begins only from a named evidence gap, bounded query plan, budget, and stop condition.
Before reporting candidate intake complete, settle its ranking against the current Decision Context:
top_window, ranked_backlog, or
no_change;no_change and use it only for standard-value or
substantially duplicative materials;If ranking apply, readback, projection, or rollback readiness fails, repair or roll back before reporting the intake as complete.
A ranked entry must represent one independently sortable reading or action unit, not a display bucket.
Splitting is semantic, not mechanical. Group materials only when they jointly support one decision or learning outcome; separate materials whose value, urgency, reader action, or evidence maturity differs.
Rerank from revision-bound Decision Context evidence.
plan_material_single_move with the complete ranked set, including backlog.
Preserve all other relative order and use its exact interval bounds in the
existing proposal. Recompute that preview before apply; do not relax the
constraints of a historical proposal.plan_material_wave_rerank (or
loopx material-lifecycle plan-wave-rerank --input <path-or-->) over the
complete ranked set before building material_rerank_proposal_v0, enforcing
all three ranking quality rules:
imminent_gate, active_sprint, foundational_track, and
orthogonal_backlog with cost-adjusted VOI (quick_slice /1.0,
focused_study /2.0, deep_monograph /6.0). Automatically stratify
unsliced deep_monograph items (scoped_slice_ready=False) declared in
imminent_gate or active_sprint into foundational_track so multi-week
courses or monographs never block quick high-VOI slices unless an explicit
bounded chapter/section slice is ready.required_load_bearing_refs are already ingested and ranked before
planning a rerank (settle candidate intake first when any load-bearing
primary is missing or rank=None), enforce topological primary-over-secondary
precedence (primary_source_refs strictly ahead of secondary commentary),
and keep active-deliverable load-bearing primaries
(blocks_active_deliverable=True, evidence_role="load_bearing_primary")
inside top_window_size.orthogonal_backlog and
orthogonal_signal items inside 1..top_window_size via
max_orthogonal_in_top_window (default 0) whenever non-orthogonal
ranked items remain below top_window_size, shifting excess general-trend
items cleanly to the start of ranked_backlog while keeping untouched
backlog items in their exact original relative order.material_rerank_receipt_chunks when a transition affects more than 100
refs. Build all existing receipts before CAS and persist all after readback;
verify complete coverage under the same proposal, revisions and owner gate.The managed catalog remains authority, but operators need a readable view. Build that view from exact-read-backed presentation records:
LoopX owns validation, rendering, and the content-free receipt contract. The project adapter owns source parsing and supplies already-promoted display records. One-time legacy parsing and migration scripts remain project-local.
Preview first. Apply only when all are true:
After apply, read the destination back and write an audited receipt. On any mismatch, restore the previous authority and record the blocker. Never report "migration complete" from a preparation plan alone.
A project-specific skill or AGENTS.md may define:
It should reference this skill instead of duplicating generic migration, ranking, rebuild, apply, or rollback rules. Project rules may tighten these invariants but must not weaken them.
A complete material operation reports:
Stop without mutating the source when:
huangruiteng/CS-Notes
feda44a
If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.