[yyjeqhc]

Why

Twenty Runner-backed Project read tools repeat the same visibility, scope, authority, effect and idempotency declaration. Runner admission, detached recovery and Registry reconciliation also repeat the same validation-step and metadata consistency rules. Keeping copies makes a contract change easy to apply unevenly.

Changes

• Share the fixed read-only declaration for 8 LSP, 6 Git and 6 file tools. Each declaration still explicitly selects its audit policy, category, Runner capability, path hint and Session evidence policy; write/execution tools retain their existing declarations.
• Share Artifact ranged-read continuation argument constraints while retaining each operation's selector, description, required fields and SHA-256 fence.
• Put bounded unique validation-step names and optional metadata/name agreement on the existing ShellJobContext type. Keep operation-specific admission, authority checks, check order and error messages at their current boundaries.
• Use the existing ExecutionPurpose parser instead of three copied eight-value allowlists.
• Teach the existing source-based discovery catalog and corpus checks to recognize the shared constructor. Corpus scope and visibility are read from its actual Rust body rather than duplicated as parser constants.

No tool, Runner wire field, persisted field, permission or output contract is removed.

Size

Counts include comments and blank lines in the Git diff; production/test classification follows dedicated test paths and cfg(test) scopes.

Scope	Added	Deleted	Net reduction
Production (Rust and Python)	170	457	287
Tests (Rust and Python)	64	5	-59
Total	234	462	228

The installed MSI cloc 0.6.2 reports 257 fewer Rust code lines across the changed files (5,881 → 5,624). The Rust-only physical reduction is 246 lines; the Python corpus/parser support adds 18 net lines. No test scenario or assertion was removed.

Validation

Performed on MSI Windows against base 66dc970:

• Captured every ToolDefinition and output schema with experimental-code-mode and workspace-checkpoints enabled before and after: all 193 entries exactly equal.
• Core validation-context tests: 2 passed.
• Tool contracts, all features: 328 passed, 1 ignored.
• Registry reconciliation tests: 83 passed.
• Runner recovery-context tests: 7 passed.
• Python discovery/catalog/corpus/selection tests: 33 passed, 1 POSIX-only skip.
• Final cargo fmt --all -- --check and git diff --check.

The temporary inventory example was removed after comparison. No full-workspace or cross-platform process suite was run.