seam(a2ml): 0-AI-MANIFEST.a2ml is template prose with rsr-template-repo vestiges — not actually an a2ml manifest

[hyperpolymath]

Diagnosis

Seam-audit 2026-06-01. Compare echidna and echidnabot's root manifest:

echidna/0-AI-MANIFEST.a2ml (correct shape — TOML/s-expr a2ml):
```
[manifest]
version = "1.1"
project = "echidna"
type = "neurosymbolic-theorem-prover"
full-version = "1.6.0"
license = "MPL-2.0"
…
```

echidnabot/0-AI-MANIFEST.a2ml (markdown prose, with leftover template scaffolding):
```

AI Assistant Instructions

Repository Focus

• `rsr-template-repo` is treated as a Rhodium Standard Repository…
• Prefer to keep generated files out of source control…

Workflow

1. Inspect `.machines_readable/6scm/STATE.scm` for blockers…
   ```

Two compounding bugs in the echidnabot file:

1. Not a2ml at all. The file claims .a2ml extension but contains markdown prose. Per the new a2ml-validate (which now accepts both TOML and s-expression forms — see standards memory feedback_a2ml_validate_sexpr_form_supported), the validator expects manifest syntax, not prose.

2. Wrong repo identity. Content talks about "rsr-template-repo" and ".machines_readable/6scm/STATE.scm" — both are template vestiges. Real shape used in this repo is .machine_readable/6a2/STATE.a2ml (no s, 6a2 not 6scm).

Suggested fix shape

Mirror echidna's [manifest] / [context-tiers] / [agentic-pointers] layout:

```
[manifest]
version = "1.1"
project = "echidnabot"
type = "tier-1-verifier-bot"
full-version = "0.1.0" # tracks Cargo.toml [package].version
license = "MPL-2.0"
author = "Jonathan D.A. Jewell j.d.a.jewell@open.ac.uk"
repo = "https://github.com/hyperpolymath/echidnabot"

[context-tiers]
tier_0 = ["0-AI-MANIFEST.a2ml", ".claude/CLAUDE.md"]
tier_1 = ["README.md", "EXPLAINME.adoc", "Cargo.toml"]
tier_2 = [".machine_readable/6a2/STATE.a2ml", ".machine_readable/6a2/AGENTIC.a2ml"]
```

Concrete one-PR change

• Rewrite 0-AI-MANIFEST.a2ml against the echidna template.
• Verify with the a2ml-validate pre-commit hook (or a2ml-validate-action in CI if wired).

Don't-touch boundaries respected

• The .machine_readable/6a2/ directory itself is fine — just the root sibling file needs replacing.

This is small but did not feel safe to action in the seam-audit sweep without owner sign-off on the schema shape, hence diagnosis-only.

[hyperpolymath]

Resolved by #49 — no separate PR needed.

Verification

PR #49's diff for 0-AI-MANIFEST.a2ml (gh pr diff 49 -- 0-AI-MANIFEST.a2ml, 40 add / 11 del) replaces the entire markdown prose body with a valid TOML-flavoured a2ml manifest matching exactly the shape this issue's "Suggested fix shape" prescribed:

• [manifest] block with project = "echidnabot", license = "MPL-2.0", estate = "hyperpolymath", fleet, etc.
• [entry-points] block listing human-readable / ai-instructions / ai-state (pointing at .machine_readable/6a2/STATE.a2ml, META.a2ml, AGENTIC.a2ml, ECOSYSTEM.a2ml) / bot-directives / abi-ffi-surface (covers Rust src/abi/ and Zig ffi/zig/)
• [workflow] block with inspect-first / respect / self-directive keys
• [promises] and [history] blocks
• SPDX header restored at top

Both compounding bugs from the diagnosis are fixed:

1. Template-prose body → valid a2ml syntax (TOML form, recognised by a2ml-validate per feedback_a2ml_validate_sexpr_form_supported).
2. rsr-template-repo + .machines_readable/6scm/ (typo'd s + Scheme-era dir) vestiges → all references now point at the correct .machine_readable/6a2/*.a2ml. The [history] block even documents the 2026-04-17 Scheme → a2ml migration that left the stale prose behind.

Confirmed by gh api search/code?q=repo:hyperpolymath/echidnabot+machines_readable that the only file in the repo that ever contained the machines_readable typo is 0-AI-MANIFEST.a2ml itself, which #49 rewrites.

PR #49 is already auto-merge-armed (SQUASH); current BLOCKED state is due to 4 pre-existing red checks (PR (address), two governance jobs, openssf-compliance) that are baseline-red on main, not regressions from this PR.

Closing as resolved-by-#49. If the merge of #49 somehow drops this hunk, re-open and I'll cut a follow-up PR.