fix(6854): the locating arguments that only point are declared as references, and the plugin version is minted
CI & Build / Python lint (push) Successful in 2s
CI & Build / Plugin hooks (push) Successful in 13s
CI & Build / TypeScript typecheck (push) Successful in 32s
CI & Build / Python tests (push) Successful in 1m43s
CI & Build / integration (push) Successful in 2m2s
CI & Build / Build & push image (push) Successful in 22s

The guard caught seven write-tool arguments with a locating name that are not
where the write lands:
- snippet_id on classify_shapes_by_rule and confirm_shape_proposals names the
  canon, not the rows written;
- rule_outcome, rule_misfired and judge_lesson_link record judgments.

REFERENCES lists each one with its reason, and the guard reads it.
READ_PREFIXES names the tool prefixes the fence answers without the server,
and a test holds every write tool to a name outside them.

plugin.json is minted to 2026.11.05.0242 for the PACKAGING.md row.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
This commit is contained in:
2026-11-04 21:42:14 -05:00
co-authored by Claude Opus 5.5
parent 55f0f96016
commit 16b1d68ebd
3 changed files with 53 additions and 5 deletions
+32 -1
View File
@@ -143,12 +143,43 @@ WRITE_TARGETS: dict[str, dict[str, str]] = {
# Argument names that would locate a write wherever they appear. The guard test
# requires every write tool parameter with one of these names to be declared
# above, so a tool that gains one can't write past the fence unnoticed.
# above, or listed in REFERENCES below, so a tool that gains one can't write
# past the fence unnoticed.
LOCATOR_NAMES = frozenset({
"project_id", "task_id", "note_id", "lesson_id", "snippet_id", "process_id",
"milestone_id", "system_id", "rule_id", "deleted_batch_id",
})
# Arguments with a locating name that only point at a record, with why each one
# doesn't say where the write lands.
REFERENCES: dict[str, dict[str, str]] = {
"classify_shapes_by_rule": {
"snippet_id": "the canon the shapes are judged against; the rows written are project_id's",
},
"confirm_shape_proposals": {
"snippet_id": "narrows which proposals to confirm; the rows written are project_id's",
},
"judge_lesson_link": {
"lesson_id": "a judgment on a lesson-to-rule pairing the system asked for",
"rule_id": "a judgment on a lesson-to-rule pairing the system asked for",
},
"rule_outcome": {
"rule_id": "telemetry on what a rule changed in this session's own work",
},
"rule_misfired": {
"rule_id": "telemetry on a moment that fired a rule in this session",
"project_id": "the project the misfire happened in, recorded with the report",
},
}
# Tool-name prefixes the plugin's fence answers without asking the server: every
# tool named this way reads. tests/test_write_scope.py holds the hook's copy of
# this list equal to it, and holds every write tool to a name outside it.
READ_PREFIXES = (
"get_", "list_", "search", "find_", "enter_project", "what_might_apply",
"resolve_design_system",
)
# The one write a session may make in another project: filing an issue there,
# as a new open task outside any plan of theirs.
ISSUE_STATUSES = frozenset({"", "todo"})