Skillscript

Closes Perry's spec request (thread 1f278e5e): an optional, generic,
connector-implemented author field on the skill discovery surface
plus an author filter on skill_list. Use case: "here's your
toolbox" session-start context limited to the agent's own skills.
Substrate-neutral — populating + filtering is the connector's job;
adopters with no authorship concept get graceful degradation.

A. SkillEntry.author on catalog output

buildSkillCatalog() now surfaces SkillMeta.author on each SkillEntry.
Bundled stores already track author (FilesystemSkillStore reads from
os.userInfo().username, SqliteSkillStore stores at write time);
adopter substrates that don't track authorship surface null per the
contract's graceful-degradation rule.

B. skill_list filter.author

SkillListFilter.author AND-composes with the existing filters
(audience/status/trigger_kind/domain_tags/name_prefix). The catalog
layer threads the filter through to SkillStore.query({ author: ... })
so substrates that honor it natively can narrow at the substrate
layer; the catalog ALSO applies the filter in-memory against
meta.author so substrates that ignore the filter still produce the
right result (graceful degradation).

SkillFilter.author was already a declared field on the contract;
v0.18.6 wires the catalog + MCP surface to actually thread it through.

Tests

6 new tests in v0.18.6-author-discovery.test.ts:

• Author surfaced on entries
• filter.author narrows correctly + entries carry matching author
• Empty result on no match (no error)
• AND-composes with name_prefix
• Graceful degradation: substrate ignores filter, catalog narrows
• author === null when substrate doesn't populate (mock store)

1719 tests total; typecheck clean. No LOC ceiling bump needed
(catalog edits absorbed within budget).

Docs

• mcp-server.ts skill_list tool docstring lists the new filter +
  documents graceful degradation
• docs/connector-contract-reference.md SkillStore conventions section
  adds author-field convention + filter passthrough rule

Adopter-side action (substrate-author backfill — out of scope)

Connectors backed by substrates with an existing authorship concept
need to either honor the new filter.author in their query() natively
(better performance) or rely on the catalog-layer in-memory filter
(works out of the box). Substrates with legacy data missing author tags
(e.g., skills predating identity-stamping) need data-side backfill —
substrate-author work, not a language change.