Efficiency Layer — First Candidate Skills

Reference documentation for Efficiency Layer — First Candidate Skills in Adaptive Skills.

Why these three come first

The first Efficiency Layer candidates should solve repeated operational problems without collapsing into orchestration policy.

That is why the first set is:

• task-chunking
• handoff-summary
• checkpoint-review

They improve how work is carried out, but they do not require the repository to make stronger assumptions about model runtime, memory systems, or hidden orchestration.

task-chunking

Status: published in skills/task-chunking/ as the first Efficiency Layer skill.

Purpose

Break oversized work into smaller, reviewable, and more testable slices.

Helps when

• a task keeps entering as one large block
• the first useful slice is still unclear
• dependencies are making the task feel bigger than necessary

Expected output

• smallest useful slice
• next slice boundary
• dependency notes
• stop condition for the current slice

Boundary

• If the problem is feature definition, feature-planning is still primary.
• task-chunking is transversal: it helps keep execution bounded after or alongside planning.

handoff-summary

Status: published in skills/handoff-summary/ as the second Efficiency Layer skill.

Purpose

Close one round of work with enough context for the next round without dragging the whole session forward.

Helps when

• work must continue later
• another person or agent will resume it
• the context is getting heavy and a clean restart is healthier

Expected output

• current state
• what was already proved
• what remains open
• next recommended step
• known risks or missing evidence

Boundary

• If the handoff is about macro gate, review, or cross-boundary continuity, AletheIA remains the macro layer.
• handoff-summary is the micro execution aid for ending a round cleanly.

checkpoint-review

Status: published in skills/checkpoint-review/ as the third Efficiency Layer skill.

Purpose

Insert a short, useful pause between rounds so teams do not keep pushing work forward blindly.

Helps when

• the session is growing long
• the task shape changed mid-way
• the team needs to decide whether to continue, stop, or hand off

Expected output

• what changed in this round
• what is now known
• whether the next step belongs in the same round
• whether a handoff or stop is healthier

Boundary

• If the task needs formal gate or escalation, AletheIA remains responsible for macro posture.
• checkpoint-review is the lighter operating pattern for local round discipline.

Why the others are deferred

context-compression

Useful, but closer to memory/summary systems and therefore riskier to define too early.

prompt-framing

Useful, but can overlap with general task framing, consumer onboarding, and AletheIA macro framing.

model-routing

Useful, but too close to orchestration and model strategy; it should wait until the simpler efficiency skills are better proven.