v1.0-exp pack schema.
v1.0-exp is an experimental extension of v0.1. All v0.1 fields remain valid. See Compatibility for guidance on mixing versions.Complete example
Field reference
Top-level fields
Schema version. Always
"1.0-exp" for experimental packs. Use "0.1" for packs that do not use any experimental fields.Unique slug identifier for this pack. Use kebab-case. This id is referenced by
index.json and by merge_suggestions in other packs.Example: "html-to-nextjs-migration"Declares which level this pack operates at. One of:
"task"— applies to a specific recurring task type."project"— project-wide defaults and constraints."user"— user collaboration style that travels between repos.
Human-readable title displayed when the agent reports active packs.Example:
"HTML to Next.js migration"Lifecycle status of the pack. One of:
"draft"— work in progress, not yet trusted for automatic application."approved"— reviewed and trusted.
Provenance metadata describing how this pack was created.
Signals used by the agent to decide whether this pack is relevant to the current task.
(v1.0-exp, recommended) Contextual hints that help the agent decide how well this pack fits the current repo and task. Agents treat these as soft filters — they boost or dampen scores but do not hard-disqualify a pack.
(v1.0-exp) A single scalar in
[0, 1] representing how strongly this pack encodes a proven pattern.Interpretation:0.3— weak or tentative pattern.0.7— solid pattern that has worked multiple times.0.9+— very strong, highly trusted pattern.
combined_score = f(match_score, strength).Short statements describing what this pack is trying to achieve.
Hard rules the agent must follow while this pack is active. Violations indicate the pack is being misapplied.
Soft preferences — the agent should follow these unless a strong reason not to exists.
Ordered or unordered steps describing the recommended workflow for this task type.
Known failure modes and anti-patterns to avoid when applying this pack.
Conditions to verify before considering the task complete. The agent can use these as a checklist.
(v1.0-exp) Simple usage counters and timestamps. Agents may update these after each
aep apply.(v1.0-exp) Ordered log of meaningful changes to this pack. Each entry records what changed and why.
(v1.0-exp) Hints that this pack overlaps with another and the two may be candidates for merging. Agents should not auto-merge based on these; they are cues for human review.
Supporting evidence and notes attached to this pack.
Related pages
Index schema reference
Field reference for the index.json file that tracks all packs.
Templates
Ready-to-use JSON templates for all three pack scopes.