Skip to content

askit-capability-whats-new

The first of three skills in the capability family. It answers one question and refuses the next one.

Skill Question
askit-capability-whats-new what shipped upstream since we last looked?
askit-capability-gap-analysis what does that mean for us?
askit-capability-advisor what can an author build on a given agent?

A plugin standard asserts facts about software it does not control: what a runtime loads, what a manifest may declare, which component types can ship. Two kinds of machine keep those honest, and they are not the same kind.

A watch re-checks something already written down - askit-standards-watch against a pinned specification blob, vendor-watch against pinned vendor sentences. Both answer “did a fact we depend on stop being true?”

Neither can see a capability nobody has written down yet, because a capability with no pin has nothing to fail. That is this skill’s question, and no watch can be extended to cover it.

  1. Reads a pin recording, per surface, the version last surveyed.
  2. Fetches each surface named in its static index - never a constructed URL.
  3. Classifies every release entry as capability, claim-bearing, environmental, or not relevant (counted, not listed).
  4. Confirms a capability entry against the documentation page, not the release note.
  5. Appends a dated section to the survey record - including when it found nothing.
  6. Proposes a new pin. It does not write one.

It pins a version, not a date. Every surface publishes versioned entries, so “everything after 2.1.208” is exact and re-derivable by anyone. A date pin silently loses anything published out of order and gives two readers different answers to the same question.

It is a periodic human survey, not an automated diff. Measured on 2026-08-18: one platform moved through 29 versions inside a single changelog window, and another carried 31 entries. An alarm firing weekly on entries that almost never matter trains its reader to close it unread - and then the alarm’s existence is itself false assurance. The same reasoning is why vendor-claims.json pins sentences rather than hashing pages.

It reads documentation pages, not just release feeds. The single most useful finding of its first run was announced by no release note at all: a documentation page states which component types a Codex plugin may contain, and that sentence both corroborated something previously known only by experiment and named three component types the Standard does not model.

Decide anything. It does not edit the capability matrix, a check, the Standard, a vendor claim, or a component’s frontmatter. A survey that also proposed changes would be graded on how much it found, and a survey graded on volume stops reporting “nothing material” - which is its most common and most useful result.

It also cannot see a capability the vendor shipped and documented nowhere. That gap closes only by probing, which is a different mechanism with a different cost.

Artifact Nature
foundation/surveys.md append-only dated record; never rewritten
foundation/claims/surveyed-pin.json proposed, landed by a human alongside the record
candidate claims routed to vendor-claims.json, not written by this skill

Nothing here blocks a release. An old survey means work is waiting, not that a claim expired - which is deliberately unlike a stale probe claim, whose age is its whole verification.