Skip to the rows

agentsclimarketplace

Aoa session harvest

Skill 8Dionysus/aoa-skills/skills/core/session-growth/aoa-session-harvest

Bounded agent-facing skill bundles for coding agents and humans: reviewable execution workflows composed from reusable techniques.

Install

npx -y skills add 8Dionysus/aoa-skills --skill aoa-session-harvest

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

4 stars4 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Use only when a literal ref to a closed/reviewed session packet is supplied, to extract, classify, propose promotion or branching, or identify an automation opportunity. Missing, live, or unreviewed evidence must fail closed without inventing a ref. Natural closeout preservation belongs to aoa-memo-writeback. Do not use for generic summaries, progression, direct owner writes, scheduling, mutation, or telemetry authority.

SKILL.md

5.0 KB, as published. Nobody here has run it

aoa-session-harvest

Intent

Metabolize reviewed experience without collapsing extraction, object classification, promotion, route choice, and automation readiness into one generic candidate label. Destination owners retain all acceptance authority.

Trigger boundary

Use this skill when:

an explicitly reviewed, bounded session/evidence packet contains reusable units, an isolated unit needs owner or promotion judgment, several next routes must remain visible, or a repeated manual route needs automation triage
the same extraction or promotion is explicitly requested from live, unreviewed, or unbounded material and the only valid result is blocked_unreviewed_evidence before a mode loads

Do not use this skill when:

the request is generic summarization, progression scoring, memo writeback, runtime repair, or direct owner writing, promotion, scheduling, delegation, or mutation
the request starts from ordinary work/closeout evidence and asks whether a bounded lesson should be preserved for future owner work; that first writeback judgment belongs to aoa-memo-writeback, even if the answer might later resemble a candidate classification

Inputs

reviewed evidence boundary and at least one literal supplied ref, closure/review posture, one mode intent, relevant owner constraints, risks, effects, and optional checkpoint/handoff hints that have not yet been trusted; or an attempted live/unreviewed input used only to return the pre-mode block

Outputs

one typed session-local result from references/contract.yaml, with evidence refs, owner hints, rejected nearest route, effect none, stop line, and destination-owner review requirement; or blocked_unreviewed_evidence

Procedure

Mode	Use when	Reference
`extract`	Reusable units and unresolved residue have not been isolated.	`references/extract.md`
`classify`	One unit needs its object kind, owner, and next artifact classified.	`references/classify.md`
`promote`	Exactly one isolated repeated quest-shaped unit needs a final promotion verdict.	`references/promote.md`
`branch`	Several materially different continuations must stay visible.	`references/branch.md`
`automation-opportunity`	A repeated manual route needs readiness and highest-honest-mode classification.	`references/automation-opportunity.md`

Mode: extract

Read and follow references/extract.md.

Mode: classify

Read and follow references/classify.md.

Mode: promote

Read and follow references/promote.md.

Mode: branch

Read and follow references/branch.md.

Mode: automation-opportunity

Read and follow references/automation-opportunity.md.

Read references/contract.yaml. Confirm that the request or supplied packet contains at least one literal evidence ref; never infer, normalize, or mint a missing packet/session identifier. Then inspect only enough of that exact evidence boundary to verify explicit review/closure.
If a literal ref, review/closure, or a bounded evidence source is missing, return blocked_unreviewed_evidence without loading any mode reference.
Choose exactly one mode and read only its reference to EOF.
Execute exactly one mode. Later modes are typed handoffs and require a new selection; they are not an automatic pipeline.

Contracts

candidates and refs remain session-local until an owner accepts a handoff
extraction, classification, promotion, automation, execution, progression, telemetry, and durable owner writes are separate effects
hints, receipts, counts, and route cards never become evidence or authority merely by appearing in a reviewed packet
technique records may explain lineage but are not runtime dependencies

Risks and anti-patterns

mining a live transcript, universalizing one success, or turning themes and frustration into reusable units
inventing or normalizing a packet/session ref that was not literally supplied
forcing every repeated unit into a skill or every automation candidate into a scheduler/playbook
treating quest, receipt, progression, stats, routing, KAG, or owner state as procedure output

Verification

trace every result to reviewed refs and disposition stale/cross-session hints
confirm one mode, typed output, owner, rejected nearest target, effects, and stop line
confirm no owner file, quest, memory, route, KAG, telemetry, schedule, or runtime state changed

Adaptation points

Sessions supply reviewed packets and evidence boundaries. Destination owners supply intake, promotion, automation, progression, and artifact contracts.

Keep looking

Skills are one crate of 328,083. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.