Skip to main content
← Back to list
01Issue
FeatureClosedSwamp ClubPublic
AssigneesNone

Relationships

#3304 Docs: --literal-inputs and data that never evaluates as authored (swamp-club#3205)

Opened by stack72 · 10/12/2026

swamp-club#3205 changes user-facing behaviour that the manual describes:

  • content/manual/reference/cel-expressions.md, "Evaluated and Inert Sources": add swamp model method run --literal-inputs (inert), and state that expression text arriving as data (workflow/webhook inputs, data.*, step outputs, forEach items) stays literal even when it is identical to an expression the author wrote elsewhere.
  • content/manual/reference/model-definitions.md, the method-run inputs section: document --literal-inputs (inputs reach the method verbatim, @ is not read as a file, a routed global argument holding expression text is refused, cannot combine with --last-evaluated, over --server needs a serve advertising the literal-inputs capability).
  • content/manual/reference/vaults.md, "Model Method Run Inputs Are Evaluated": point untrusted text at --literal-inputs.

Source of truth: design/enablers/expressions.md "Position and Value Provenance" and the updated swamp skill references in the swamp repo.

02Bog Flow
✓OPEN○TRIAGED○IN PROGRESS◉CLOSED

Closed

10/12/2026, 9:40:52 PM

No activity in this phase yet.

03Sludge Pulse
Editable. Press Enter to edit.

stack72 commented 10/12/2026, 9:40:52 PM

Documented in the manual by swamp-club/swamp-club#1324 (merged).

Sign in to post a ripple.