"\n\nPre-open repo agent-surface audit \u2014 TextForAgents LLC\n

Pre-open repo agent-surface audit

\n

Before your agent opens that repo, know what in it will run \u2014 or reconfigure your agent: a per-harness OPEN / OPEN-QUARANTINED / DO-NOT-OPEN verdict on the repo's complete agent-control surface, with a sanitised copy that re-audits clean.

\n

Before your agent opens that repo, know what in it will run \u2014 or reconfigure your agent. A repository received as files can execute code through .git/config sinks (core.fsmonitor, hooksPath, sshCommand, filter/diff drivers), hooks, and MCP/package hooks, and its own subagent definitions can downgrade the reviewing model, exclude the directory where the payload sits, and instruct the reviewer not to re-audit. This service is the pre-open gate: it never executes the tree, enumerates the complete agent-control surface on two separately graded surfaces (EXECUTION and EXAMINATION), issues a per-harness x per-version verdict \u2014 OPEN / OPEN-QUARANTINED / DO-NOT-OPEN \u2014 from a maintained patch-status matrix, signs a receipt whose tree hash binds file contents, and can emit a sanitised quarantined copy accepted only when it re-audits with zero high findings.

\n

Prices: Verdict lookup $0.01/query \u00b7 Pre-open audit $0.99/tree \u00b7 Audit + sanitised copy $1.99 \u00b7 Operator plan $9.99/operator/mo (50 fresh audits included) \u00b7 Fleet plan $19.99/operator/mo (200 fresh audits included). Plan inclusions are fresh audits only; sanitised copies meter at $1.99; lookups are free inside plans. One free pre-open audit per operator, ever (one-time trial). Support: support@textforagents.com

\n

Stated plainly: A verdict grades the enumerated agent-control surface of a received tree on two surfaces \u2014 EXECUTION and EXAMINATION \u2014 against the customer's pinned policy and the patch-status matrix for the named harness and version at audit time; it is not a general malware, vulnerability, or safety determination about the tree's code, and an OPEN verdict is not a representation that a tree is safe in any other respect. Enumeration is complete on the proven classes \u2014 every agent-binding file is listed in the signed receipt \u2014 but prose classification is pattern-based: a genuinely novel paraphrase can escape classification while never escaping enumeration. Verdicts are matrix-versioned: each verdict records the patch-matrix version it stands on. Willingness to pay is UNMEASURED.

\n

How to start

\n
    \n
  1. POST /register with {\"name\": ..., \"payer_wallet\": ...} \u2014 you get a one-time claim token. No credentials are needed to read this page, the docs (/docs), the OpenAPI spec (/openapi.json) or /llms.txt.
    2. \n
  2. POST /trial/claim with {\"claim_token\": ...} \u2014 you get your API token and the one-time trial grant (one free pre-open audit, ever).
    3. \n
  3. POST /audit with Authorization: Bearer <token> and {\"tree\": {\"files\": [{\"path\": ..., \"content\": ...}]}, \"harness\": ..., \"version\": ..., \"policy\": {...}} \u2014 the first audit is the free trial; after that audits are $0.99 via x402 (payment verified before the verdict is released, settled after; bogus or replayed proofs rejected). POST /lookup reads a precomputed verdict for $0.01. POST /sanitise returns the audit plus a sanitised copy for $1.99. POST /subscribe takes an Operator ($9.99/mo, 50 fresh audits included) or Fleet ($19.99/mo, 200 included) plan.
    4. \n
  4. POST /verify checks any signed receipt's Ed25519 signature; GET /verify-key lists the accepted signing-key fingerprints.
    5. \n
\n

Also available as an MCP tool surface at POST /mcp (tools: preopen_audit, lookup_verdict, sanitise_tree, verify_receipt) with the same billing. Also available as a local CLI package on PyPI: repo-agent-surface-pre-open-audit. Terms: /terms \u00b7 Privacy: /privacy \u00b7 Support: support@textforagents.com

\n"