chore: rename agent_instructions to bot_directives for spec alignment

Author: hyperpolymath

…_readable/agent_instructions/README.adoc …hine_readable/bot_directives/README.adoc.machine_readable/agent_instructions/README.adoc renamed to .machine_readable/bot_directives/README.adoc .machine_readable/agent_instructions/README.adoc renamed to .machine_readable/bot_directives/README.adoc

@@ -33,7 +33,7 @@ Methodology-aware configuration for AI agents. Read by any AI agent
 == Relationship to Other Files
 
 * `AGENTIC.a2ml` says WHAT agents can do (permissions, gating)
-* `agent_instructions/` says HOW agents should work (methodology)
+* `bot_directives/` says HOW agents should work (methodology)
 * `bot_directives/` says what the gitbot-fleet does (fleet-specific)
 * `CLAUDE.md` says how Claude specifically should work (Claude-specific)
 

…eadable/agent_instructions/coverage.a2ml …ne_readable/bot_directives/coverage.a2ml.machine_readable/agent_instructions/coverage.a2ml renamed to .machine_readable/bot_directives/coverage.a2ml .machine_readable/agent_instructions/coverage.a2ml renamed to .machine_readable/bot_directives/coverage.a2ml

@@ -15,5 +15,5 @@ Disallow-Generation: yes
 # For AI agent integration instructions, see:
 #   0-AI-MANIFEST.a2ml (lexicographic-first AI entry point)
 #   .machine_readable/INTENT.contractile (purpose + ecosystem)
-#   .machine_readable/agent_instructions/methodology.a2ml (mode/budget/invariants)
+#   .machine_readable/bot_directives/methodology.a2ml (mode/budget/invariants)
 #   docs/audits/ (drift-reconciliation audit log)

…ne_readable/agent_instructions/debt.a2ml …achine_readable/bot_directives/debt.a2ml.machine_readable/agent_instructions/debt.a2ml renamed to .machine_readable/bot_directives/debt.a2ml .machine_readable/agent_instructions/debt.a2ml renamed to .machine_readable/bot_directives/debt.a2ml

@@ -3,7 +3,7 @@
 #
 # Lexicographic-first machine-readable file so any agent that does a
 # default-sorted directory walk discovers this first. Points at the
-# canonical contractile family and the agent_instructions/ tree.
+# canonical contractile family and the bot_directives/ tree.
 
 [manifest]
 project = "valence-shell"
@@ -22,9 +22,9 @@ contractile-purpose       = ".machine_readable/INTENT.contractile"
 contractile-must          = ".machine_readable/MUST.contractile"
 contractile-trust         = ".machine_readable/TRUST.contractile"
 contractile-adjust        = ".machine_readable/ADJUST.contractile"
-methodology               = ".machine_readable/agent_instructions/methodology.a2ml"
-coverage                  = ".machine_readable/agent_instructions/coverage.a2ml"
-debt                      = ".machine_readable/agent_instructions/debt.a2ml"
+methodology               = ".machine_readable/bot_directives/methodology.a2ml"
+coverage                  = ".machine_readable/bot_directives/coverage.a2ml"
+debt                      = ".machine_readable/bot_directives/debt.a2ml"
 human-claude-handbook     = "CLAUDE.md"
 human-readme              = "README.adoc"
 human-explainme           = "EXPLAINME.adoc"

…able/agent_instructions/methodology.a2ml …readable/bot_directives/methodology.a2ml.machine_readable/agent_instructions/methodology.a2ml renamed to .machine_readable/bot_directives/methodology.a2ml .machine_readable/agent_instructions/methodology.a2ml renamed to .machine_readable/bot_directives/methodology.a2ml

@@ -49,7 +49,7 @@ Cross-validation. The six systems sit on different logical foundations:
 If all six prove the same theorem, it is *extremely* unlikely that all six
 modelled the problem incorrectly in the same way. This is the same pattern
 seL4 and CompCert use, and it is the central technical commitment of the
-project — see `INTENT.contractile` and `.machine_readable/agent_instructions/methodology.a2ml`
+project — see `INTENT.contractile` and `.machine_readable/bot_directives/methodology.a2ml`
 for the architectural-invariant statement.
 
 == How do I trust the proofs?

.well-known/ai.txt

@@ -108,7 +108,7 @@ Key invariants that must never be violated:
 * No new TypeScript / Python / Go files; no npm/bun/yarn/pnpm dependencies (Deno only)
 * No `believe_me` or `assert_total` (Idris2); no new `Admitted.` (Coq); no `sorry` (Lean); no `unsafeCoerce` (Haskell); no `Obj.magic` (OCaml)
 * No `unsafe {}` blocks without an inline safety comment (Rust)
-* No template placeholders (`{{...}}`, `<...>`) in shipped files — see `.machine_readable/agent_instructions/methodology.a2ml`
+* No template placeholders (`{{...}}`, `<...>`) in shipped files — see `.machine_readable/bot_directives/methodology.a2ml`
 * All commits MUST be GPG-signed
 * Full machine-checked list lives at `.machine_readable/MUST.contractile`; K9 validators in `contractiles/k9/` enforce it on every PR
 

0-AI-MANIFEST.a2ml

@@ -176,7 +176,7 @@ audit-log baseline; future audits should append rather than rewrite.
   deps + explicit "no known downstream consumers at v0.9.0" rather than
   fabricated names.
 
-* `.machine_readable/agent_instructions/methodology.a2ml` —
+* `.machine_readable/bot_directives/methodology.a2ml` —
   `PROJECT_UNIQUE_STRENGTH` placeholder filled with the polyglot-6
   cross-foundation argument and the divergence-from-siblings contrast
   (vs. echidna dyadic-Coq-only, proven single-Idris2, ephapax four-layer-Coq);
@@ -327,7 +327,7 @@ This audit followed the verisimdb#81 shape:
   vs `.md`) and consolidate on the estate-convention winner.
 . Walk the open-issue tail (#41–#46, #49, #51) and capture each as
   audit-table input.
-. Inspect sibling-repo `.machine_readable/agent_instructions/methodology.a2ml`
+. Inspect sibling-repo `.machine_readable/bot_directives/methodology.a2ml`
   (echidna, proven, ephapax) to ground the divergence-statement in
   `methodology.a2ml`.