# Phase 06 — Security Findings This directory holds per-finding remediation detail documents produced after the **internal self-audit** (`docs/security/plugin-internal-findings.md`) and after the **external pen-test** (Cure53 / SySS / secfault / etc., per `docs/security/plugin-audit-rfp.md`). ## Folder structure ``` security-findings/ ├── README.md # this file ├── F-001-capability-consent.md # one file per Critical/High finding ├── F-002-per-instruction-fuel.md ├── F-003-audit-hash-chain.md ├── ... └── external/ └── 2026-MM-DD--F-.md # external-audit findings ``` ## When to create a file here Create a per-finding file for **every Critical and High** finding from either source. Medium / Low findings can be tracked as one-line entries in the parent inventory (`docs/security/plugin-internal-findings.md` or the external sanitised report) until they are scheduled for remediation, at which point they get promoted to a file here. A finding file must contain: 1. **ID + severity** — `F-NNN` (internal) or `-` (external). Severity = Critical/High/Medium/Low + CVSS or DREAD if the auditor provided. 2. **Asset** — link to the matching `A-NN` in `docs/security/plugin-threat-model.md`. 3. **Description** — what's wrong, what an attacker can do, business impact. 4. **Reproduction** — minimal repro steps the auditor / engineer can follow. 5. **Remediation plan** — design + concrete file/line targets. 6. **Status** — `open`, `in-progress`, `remediated`, `accepted-residual` (with operator sign-off), or `wont-fix` (with explicit justification + acceptance). 7. **Remediation commits** — list of SHAs once landed. 8. **Re-test result** — auditor's re-test verdict for external findings, or the responsible engineer's signed-off internal review for internal findings. ## Template ```markdown # F-NNN: **Severity:** Critical | High | Medium | Low **Asset:** [A-NN](../../../security/plugin-threat-model.md#1-asset-catalogue) **Source:** internal self-audit (2026-MM-DD) | **Status:** open | in-progress | remediated | accepted-residual | wont-fix ## Description … ## Reproduction … ## Remediation plan … ## Remediation commits | SHA | Summary | |-----|---------| ## Re-test result … ``` ## Lifecycle 1. **Open.** Finding is recorded here as a stub with description + reproduction. 2. **In-progress.** A PR or commit chain is targeting it; the file lists the WIP branch / commits. 3. **Remediated.** All commits landed on `main`; re-test scheduled. 4. **Re-tested.** Auditor (external) or self-audit (internal) confirms the gap is closed. Status updated to `remediated`. 5. **Sign-off** (Critical/High only). The operator and lead auditor co-sign `docs/security/plugin-runtime-signoff.md` — only after **all** Critical and High items reach `remediated` (or `accepted-residual` with co-signature). ## Cross-references - `docs/security/plugin-threat-model.md` — the threat-model asset catalogue (`A-NN`). - `docs/security/plugin-attack-surface.md` — the inventory of host bindings / DB tables / routes. - `docs/security/plugin-internal-findings.md` — the running list of internal-self-audit findings. - `docs/security/plugin-audit-rfp.md` — the external-audit scope of work. - `docs/security/plugin-runtime-signoff.md` — the joint sign-off the production rollout gate. - `docs/improvement-plan/BLOCKERS.md` § B-006 — the operator-procurement blocker for the external audit.