v0.4.0: self-build safety — REPL TDD + literate prose
Some checks failed
Deploy (Gitea) / deploy (push) Failing after 2s
Some checks failed
Deploy (Gitea) / deploy (push) Failing after 2s
RED proofs (pre-v0.4.0): - dispatcher-check-secret-path 'core-loop-reason.org' → NIL (unprotected) - dispatcher-check-core-path function does not exist - Write to core file passes through gate unchanged - test-self-build-core-protection does not exist - Dispatcher suite: 19/0 GREEN proofs (v0.4.0): - dispatcher-check-core-path: T for core-*.org/lisp, NIL for others - SELF_BUILD_MODE=true: core write → :approval-required Flight Plan - SELF_BUILD_MODE=false (default): writes pass through - Dispatcher suite: 24/0 (new test-self-build-core-protection) Prose: - New 'Self-Build Safety Boundary' section: explains thin harness/fat skills corollary, regex-based core-* detection, Flight Plan vs LOG blocking, SELF_BUILD_MODE env var semantics.
This commit is contained in:
@@ -159,6 +159,19 @@ Destructive and injection patterns that are blocked in shell commands. Covers ~r
|
||||
(cl-ppcre:scan regex path)))
|
||||
|
||||
#+end_src
|
||||
|
||||
** Self-Build Safety Boundary (v0.4.0)
|
||||
|
||||
The Dispatcher now protects the core pipeline from unapproved modification. This is the operational realization of "thin harness, fat skills" — the harness is thin enough for a human to audit, and the Dispatcher ensures it stays that way.
|
||||
|
||||
The ~core-*~ files implement the Perceive-Reason-Act cycle, the Merkle-tree memory, the skill engine loader, and the Dispatcher gate stack itself. If the agent (or a hallucination) modifies these files, the agent loses its ability to reason about and fix the corruption. The Dispatcher blocks any file write or shell command targeting ~core-*.org~ or ~core-*.lisp~ — detected by ~dispatcher-check-core-path~ using direct regex matching (~core-.*\.(org|lisp)~).
|
||||
|
||||
Unlike secret path protection (Vector 2), which produces a hard ~:LOG~ block, core file writes produce a ~:approval-required~ Flight Plan (Vector 2b). The human reviews the proposed core change in an Org buffer before approving — the same mechanism that governs shell commands and network exfiltration.
|
||||
|
||||
The ~SELF_BUILD_MODE~ env var controls this protection:
|
||||
- ~SELF_BUILD_MODE=true~ (default ~false~): core path protection active — writes require HITL approval
|
||||
- ~SELF_BUILD_MODE=false~: protection disabled — useful during development when the human is manually editing core files
|
||||
|
||||
** dispatcher-check-core-path
|
||||
;; REPL-VERIFIED: 2026-05-06T18:00:00
|
||||
#+begin_src lisp
|
||||
|
||||
Reference in New Issue
Block a user