SkillPackage
https://litlfred.github.io/folio-assistant/cat-harness/package/conduct-core
JSON-LD · in graph cat-harness.jsonld ·
| name | conduct-core |
|---|---|
| version | 0.1.0 |
| description | How an agent conducts itself in this harness: interaction modality and the order of a question, confirmation and its waivers, deletion only on confirmation, agent memory, communication language, untrusted input and unverified constraints, retries, where a thing goes, and getting started. Moved out of folio-core into the conduct topic (bean 9umr, owner ruling 2026-09-30); security is its sibling in the same topic. |
| path | skills/conduct/conduct-core |
| hasManifest | true |
| declaresSkill | Subagent memory — what a durable entry may claim https://litlfred.github.io/folio-assistant/cat-harness/skill/agent-memory Agent permissions — asked for, explained, granted by a person, recorded https://litlfred.github.io/folio-assistant/cat-harness/skill/agent-permissions Ask the person https://litlfred.github.io/folio-assistant/cat-harness/skill/ask-the-person The language you communicate in https://litlfred.github.io/folio-assistant/cat-harness/skill/communication-language A confirmation can be waived — by the person who is owed it https://litlfred.github.io/folio-assistant/cat-harness/skill/confirmation-waiver /continual-progress — trackable, always-PR'd, continually-committed work https://litlfred.github.io/folio-assistant/cat-harness/skill/continual-progress The comparison goes BEFORE the question, not inside the options https://litlfred.github.io/folio-assistant/cat-harness/skill/decision-comparison Deletion requires explicit confirmation https://litlfred.github.io/folio-assistant/cat-harness/skill/deletion-requires-confirmation /getting-started — what did they actually ask for? https://litlfred.github.io/folio-assistant/cat-harness/skill/getting-started Hand back to the discussion https://litlfred.github.io/folio-assistant/cat-harness/skill/hand-back-to-discussion Harness projections: CI writes them, the owner consents to every hand edit https://litlfred.github.io/folio-assistant/cat-harness/skill/harness-projections /interaction-modality — ask in a form the person can answer https://litlfred.github.io/folio-assistant/cat-harness/skill/interaction-modality Opening brief — brief the topic before you touch anything https://litlfred.github.io/folio-assistant/cat-harness/skill/opening-brief /repo-conversion — folio-assistant over a repo that already has a life https://litlfred.github.io/folio-assistant/cat-harness/skill/repo-conversion consulted: reference material nobody performs — a PRINCIPLE, not a procedure: every heading is an argument ("Retrying a definitive answer is not caution", "Jitter is not decoration here"), and it names no actor and no trigger. It is read when writing retry code, never performed as a step. https://litlfred.github.io/folio-assistant/cat-harness/skill/retry-backoff symbiotic-interaction — the three registers of author input https://litlfred.github.io/folio-assistant/cat-harness/skill/symbiotic-interaction Untrusted input — the value must never become program text https://litlfred.github.io/folio-assistant/cat-harness/skill/untrusted-input consulted: reference material nobody performs — a DISCIPLINE. "Never encode a constraint you have not verified" is a rule that governs other work; its headings are the argument and the worked case. Nobody performs it as a task. https://litlfred.github.io/folio-assistant/cat-harness/skill/unverified-constraints Where does this go? — the question, then the skill that answers it https://litlfred.github.io/folio-assistant/cat-harness/skill/where-does-this-go Keep the context of your work https://litlfred.github.io/folio-assistant/cat-harness/skill/work-context |
| providesCapability | |
| requiresCapability | |
| inSubgraph | skills https://litlfred.github.io/folio-assistant/cat-harness/directory/skills |