← AEI Scan on Prosights.ai. This line is not part of the report: the website adds it.
This is the report AEI Scan 0.1.0 wrote on 2026-11-16 for the repository of Kyberna, the product Prosights builds. The coding agent is Claude Code. Prosights ran the scan and gave the answers. On this page the names of single rule, skill and agent files, of the worktree folders and of 1 tool are replaced by a pattern and a count; every other line is as the tool wrote it. 81 of the 97 checks come from the process used to build Kyberna, so read this as an example of a report, not as validation of the index.
Your result: AEI 79, Level 1.
What the level means: The practice is written down; nothing checks that the agent follows it.
| # | Dimension | Put in place | Where, or who confirms it |
|---|---|---|---|
| 1 | Safety controls | In every hook on PreToolUse, set onFailure to block. | .claude/settings.json, .claude/settings.local.json, ~/.claude/settings.json |
| 2 | Review and merge | Add a CODEOWNERS file. | 1 of: .github/CODEOWNERS, CODEOWNERS, docs/CODEOWNERS |
| 3 | Review and merge | Add an automated review that runs on every change before it merges and cannot be skipped. | Then show it to the assessor |
| 4 | Memory and continuity | Have each agent record that it read the carried-forward lessons before its work. | Then show it to the assessor |
| 5 | Measurement | Measure the merge rate and the revert rate of agent changes each quarter, against your own baseline. | Then show it to the assessor |
These are the first 5 of 10 actions. The appendix lists all of them, with the check behind each.
95 of 97 checks were graded: 33 by the scan, 16 by your survey answers and 46 by what a person was shown. The answers are your own, and nobody checked them unless an assessor asked for the artifact. The scan shows that a control is present, not that it works.
2 checks could not be graded, because the question was not answered or the scan could not read the evidence: 12.1, 12.4. Answer the questions and run the tool again to raise the coverage.
The score is the share of the weighted evidence that was found; it is not a percentile. The claim gap counts the checks where your answer was yes and the scan found nothing.
| # | Dimension | Level | Highest level possible | Score | Checks graded |
|---|---|---|---|---|---|
| 1 | Requirements and recording | 3, Enforced | 3 | 100 | 4 of 4 |
| 2 | Decisions | 2, Checked | 2 | 100 | 3 of 3 |
| 3 | Expected behavior and design | 2, Checked | 2 | 100 | 2 of 2 |
| 4 | Agent briefing and model policy | 3, Enforced | 3 | 100 | 6 of 6 |
| 5 | Build | 3, Enforced | 3 | 100 | 4 of 4 |
| 6 | Test and evidence | 2, Checked | 4 | 56 | 6 of 6 |
| 7 | Review and merge | 1, Written | 4 | 47 | 7 of 7 |
| 8 | Release | 4, Measured | 4 | 100 | 3 of 3 |
| 9 | Safety controls | 4, Measured | 4 | 86 | 8 of 8 |
| 10 | Memory and continuity | 1, Written | 2 | 78 | 6 of 6 |
| 11 | Measurement | 1, Written | 4 | 53 | 5 of 5 |
| 12 | Governance | 1, Written | 3 | 25 | 2 of 4 |
| 13 | Token and cost | 3, Enforced | 4 | 88 | 12 of 12 |
| 14 | Harness depth | 3, Enforced | 3 | 100 | 7 of 7 |
| 15 | Process guards | 3, Enforced | 3 | 100 | 11 of 11 |
| 16 | Model conduct | 1, Written | 4 | 27 | 9 of 9 |
Select the title of a part of the appendix to open it. Every part prints open.
At most 10, in this order: safety controls first, then whatever holds each dimension at its level, starting with the lowest. An action a scan can confirm comes before one a person must confirm.
| # | Check | Put in place | Where, or who confirms it |
|---|---|---|---|
| 1 | 9.5 (level 3) | In every hook on PreToolUse, set onFailure to block. | .claude/settings.json, .claude/settings.local.json, ~/.claude/settings.json |
| 2 | 7.2 (level 2) | Add a CODEOWNERS file. | 1 of: .github/CODEOWNERS, CODEOWNERS, docs/CODEOWNERS |
| 3 | 7.3 (level 2) | Add an automated review that runs on every change before it merges and cannot be skipped. | Then show it to the assessor |
| 4 | 10.6 (level 2) | Have each agent record that it read the carried-forward lessons before its work. | Then show it to the assessor |
| 5 | 11.2 (level 4) | Measure the merge rate and the revert rate of agent changes each quarter, against your own baseline. | Then show it to the assessor |
| 6 | 11.4 (level 4) | Keep 1 quality trend (warnings, complexity or failures) that covers at least 3 months. | Then show it to the assessor |
| 7 | 12.2 (level 3) | Put an organization-level policy for the agent in force on developers' machines. | 1 of: /Library/Application Support/ClaudeCode/managed-settings.json, /Library/Application Support/ClaudeCode/managed-settings.d/*.json |
| 8 | 16.4 (level 2) | Register a hook on the event PermissionDenied. | .claude/settings.json, .claude/settings.local.json, ~/.claude/settings.json |
| 9 | 6.3 (level 3) | Register a hook on the event Stop. | .claude/settings.json, .claude/settings.local.json, ~/.claude/settings.json |
| 10 | 13.12 (level 4) | Set env.CLAUDE_CODE_ENABLE_TELEMETRY to 1. | ~/.claude/settings.json |
The scan grades these checks too, and they are open. Every file name, key and hook event is the one the vendor documents for Claude Code. Where the vendor documents none, the row names no file.
| Check | Level | Result | Put in place | Where |
|---|---|---|---|---|
| 16.7 | 3 | Fail | Register a hook on the event SubagentStop or TaskCompleted. | .claude/settings.json, .claude/settings.local.json, ~/.claude/settings.json |
Every practice that passed, sorted by the kind of control behind it. A written rule depends on the agent choosing to follow it; a checked rule is caught after the agent has acted; an enforced rule is stopped before.
| Kind | Check | What is in place | Graded by | Where it was found |
|---|---|---|---|---|
| Written | 1.1 | Work items are written down before an agent starts on them. | survey answer | |
| Written | 2.1 | Decision records are kept in a folder of the repository. | scan | docs/decisions/ |
| Written | 2.3 | It is written which choices a person makes and which an agent may make. | survey answer | |
| Written | 3.1 | The expected behaviors of a change are written before the build. | shown to a person | |
| Written | 4.1 | The agent has an instruction file. | scan | CLAUDE.md; ~/.claude/CLAUDE.md |
| Written | 4.2 | The instruction files cover at least 4 of these 6 areas: commands, testing, project structure, code style, git workflow, boundaries. | scan | CLAUDE.md; ~/.claude/CLAUDE.md |
| Written | 4.3 | Rule files apply only to some paths. | scan | .claude/rules/*.md (17 files) |
| Written | 4.4 | Written procedures or agent definitions exist for the agent to load. | scan | .claude/skills/*/SKILL.md (2 files); .claude/agents/*.md (2 files) |
| Written | 5.1 | Each agent works in its own checkout or environment. | survey answer | |
| Written | 5.4 | The size of 1 agent task has a written limit. | survey answer | |
| Written | 6.1 | The instruction files say how tests are run. | scan | CLAUDE.md; ~/.claude/CLAUDE.md |
| Written | 7.1 | A reviewer other than its author reviews every agent change before it merges: a person, or a model run as a separate reviewer. | survey answer | |
| Written | 7.5 | It is written who may start or assign an agent. | survey answer | |
| Written | 8.1 | A person confirms each release. | survey answer | |
| Written | 9.7 | The instruction files say what the agent must never do. | scan | CLAUDE.md; ~/.claude/CLAUDE.md |
| Written | 10.1 | The agent keeps memory files for this project. | scan | ~/.claude/projects/<project>/memory/ |
| Written | 10.3 | The start-of-session procedure is written. | survey answer | |
| Written | 10.4 | Each agent writes down what it learned from its work, in a place the next agent reads. | survey answer | |
| Written | 11.5 | Agent work is measured by a named outcome. | survey answer | |
| Written | 12.3 | The rules have stable identifiers and a record of why each exists. | shown to a person | |
| Written | 13.1 | It is written which model and which reasoning effort each kind of task uses. | survey answer | |
| Written | 13.2 | Every instruction file has fewer than 200 lines. | scan | CLAUDE.md; ~/.claude/CLAUDE.md |
| Written | 13.9 | It is written that long or verbose work goes to its own agent, and that agents wait without polling. | survey answer | |
| Written | 14.2 | MCP servers are declared in a file or named in a permission rule. | scan | .claude/settings.json |
| Written | 14.4 | Agents work in worktrees the tool sets up. | scan | .claude/worktrees/* (2 folders) |
| Written | 15.4 | It is written that a new integration or tool starts with no access until it is listed. | survey answer | |
| Written | 15.8 | It is written that a change touches only what its own work item needs. | survey answer | |
| Written | 16.1 | A written record shows how each model in use has failed in the team's own work. | survey answer | |
| Written | 16.2 | Every brief to an agent states what done means. | survey answer | |
| Checked | 1.2 | A check refuses a change that is tied to no work item. | shown to a person | |
| Checked | 1.3 | A script regenerates the list of open work. | shown to a person | |
| Checked | 2.2 | A check reads the decision records at commit. | shown to a person | |
| Checked | 3.2 | A check ties each expected behavior to a test. | shown to a person | |
| Checked | 5.3 | A limit stops a build or test run that hangs. | shown to a person | |
| Checked | 6.2 | A git hook or a CI workflow is in the repository. | scan | .git/hooks/pre-commit; .git/hooks/pre-merge-commit |
| Checked | 6.6 | A gate fails on new warnings in changed files. | shown to a person | |
| Checked | 7.4 | A merge step builds and checks after each merge and stops on failure. | shown to a person | |
| Checked | 8.2 | A release gate refuses when a required item is missing. | shown to a person | |
| Checked | 9.6 | A secret scan runs before code leaves the machine. | shown to a person | |
| Checked | 10.2 | An end-of-session step refuses to finish with failing tests or unpushed work. | shown to a person | |
| Checked | 10.5 | A mechanism refuses a change whose agent recorded no lesson entry. | shown to a person | |
| Checked | 13.8 | A mechanism shortens long command output before the agent reads it. | shown to a person | |
| Checked | 14.5 | A command shows the state of each background or unattended run. | shown to a person | |
| Checked | 14.7 | The team's own hooks and guard scripts have tests. | shown to a person | |
| Checked | 15.3 | A check keeps credentials and real messages out of test fixtures. | shown to a person | |
| Checked | 15.5 | A mechanism stops a test from reaching a live account unless a person turns that on. | shown to a person | |
| Checked | 15.6 | A commit check refuses personal identifiers and authorship marks. | shown to a person | |
| Checked | 15.7 | A set of injection attempts is run against a model before it is given a channel. | shown to a person | |
| Checked | 15.9 | A guard has a single switch, and its off state is tested. | shown to a person | |
| Checked | 16.3 | Someone other than the agent checks the claims in its reports. | shown to a person | |
| Checked | 16.5 | Refused calls are logged, each with the model that made the call. | shown to a person | |
| Enforced | 1.4 | A hook runs when a prompt is submitted to the agent. | scan | .claude/settings.json: UserPromptSubmit |
| Enforced | 4.5 | The agent's model is named with a version number. | scan | .claude/settings.json (env); .claude/agents/*.md (2 files) |
| Enforced | 4.6 | A hook runs when an agent starts another agent. | scan | .claude/settings.json: PreToolUse on Agent |
| Enforced | 5.2 | A hook runs before the agent's shell commands. | scan | .claude/settings.json: PreToolUse on Bash |
| Enforced | 7.6 | A written rule says which changes need a person's approval before they merge or ship, by risk. | shown to a person | |
| Enforced | 9.1 | The agent's sandbox is on. | scan | ~/.claude/settings.json |
| Enforced | 9.2 | The agent's network access is limited to a list. | scan | ~/.claude/settings.json |
| Enforced | 9.3 | Permission rules exist, and no setting approves everything. | scan | .claude/settings.json; .claude/settings.local.json; ~/.claude/settings.json |
| Enforced | 9.4 | A hook can deny a tool call. | scan | .claude/settings.json: PreToolUse on Agent; .claude/settings.json: PreToolUse on Bash; .claude/settings.json: PreToolUse on an MCP tool; .claude/settings.json: PreToolUse on Write|Edit|NotebookEdit |
| Enforced | 13.3 | A threshold is set at which the conversation is compacted. | scan | .claude/settings.json |
| Enforced | 13.4 | The reasoning effort is set in settings or in an agent definition. | scan | ~/.claude/settings.json |
| Enforced | 13.5 | Agents that another agent starts are given their own model. | scan | .claude/settings.json; .claude/agents/*.md (2 files) |
| Enforced | 13.6 | A prompt cache lifetime is set, and caching is on. | scan | .claude/settings.json |
| Enforced | 13.10 | A spend or usage limit is set at the provider. | shown to a person | |
| Enforced | 14.1 | A hook runs before the agent writes or edits a file. | scan | .claude/settings.json: PreToolUse on Write|Edit|NotebookEdit |
| Enforced | 14.3 | An agent definition limits the tools its agent may use. | scan | .claude/agents/*.md (1 file) |
| Enforced | 14.6 | An agent runs commands on another machine only from an allow list. | shown to a person | |
| Enforced | 15.1 | Every delete goes through 1 script, and a hook refuses a delete made any other way. | shown to a person | |
| Enforced | 15.2 | A hook refuses a build or test command run outside the team's wrapper script. | shown to a person | |
| Enforced | 15.10 | A script removes an agent's workspace and refuses when its work is not merged. | shown to a person | |
| Enforced | 15.11 | No command can run outside the sandbox. | scan | ~/.claude/settings.json |
| Measured | 6.4 | The test results of each build are kept as files. | shown to a person | |
| Measured | 8.3 | The gate result of each release is recorded. | shown to a person | |
| Measured | 9.8 | A list marks each of the team's rules as enforced, checked or written only. | shown to a person | |
| Measured | 11.1 | Session records of the last 30 days are kept. | scan | ~/.claude/projects/<project>/<session>.jsonl |
| Measured | 11.3 | Usage or cost is recorded per agent or task. | shown to a person | |
| Measured | 13.7 | A report separates input tokens read from the cache from input tokens sent uncached. | shown to a person | |
| Measured | 13.11 | Tokens or cost per model are reviewed over time, and a change was made because of a review. | shown to a person |
None.
None.
You gave no counts. To fill this table, count from your own records, for each model your agents run on and for 1 period, how often it did each of these: rules broken; stopped early or asked for a permission it had; reported done when it was not; invented a reason; did work nobody asked for. Put 1 row for each model and period under model_conduct in the answers file (SCHEMA.md). Check 16.8 passes when an assessor is shown these counts for 2 periods.
The scan looked at these locations and at nothing else. It read settings, hook registrations, front matter and the headings and command lines of instruction files. It copied no file contents and no setting values into this report.
Questions answered: 16 survey, 0 on what you believe is in place, 47 shown to an assessor.
| Check | Level | Graded by | Result | What the check reads | What was found |
|---|---|---|---|---|---|
| 1.1 | 1 | survey answer | Pass | Your answer to the survey question "Are work items written down before an agent starts on them?" | the answer was yes |
| 1.2 | 2 | shown to a person | Pass | The check that refuses a change tied to no item | the artifact was shown |
| 1.3 | 2 | shown to a person | Pass | A list of open work that a script regenerates | the artifact was shown |
| 1.4 | 3 | scan | Pass | A hook on the prompt-submit event (Claude Code and Codex CLI UserPromptSubmit; VS Code userPromptSubmitted; on other platforms, shown to a person) | .claude/settings.json: UserPromptSubmit |
| 2.1 | 1 | scan | Pass | The decision-record folder you name | docs/decisions/ |
| 2.2 | 2 | shown to a person | Pass | The check that reads decision records at commit | the artifact was shown |
| 2.3 | 1 | survey answer | Pass | Your answer to the survey question "Is it written which choices a person makes and which an agent may make?" | the answer was yes |
| 3.1 | 1 | shown to a person | Pass | For 1 recent change: the expected behaviors written before the build | the artifact was shown |
| 3.2 | 2 | shown to a person | Pass | The check that ties each expected behavior to a test | the artifact was shown |
| 4.1 | 1 | scan | Pass | The agent's instruction files | CLAUDE.md; ~/.claude/CLAUDE.md |
| 4.2 | 1 | scan | Pass | Headings and command lines of the instruction files | CLAUDE.md; ~/.claude/CLAUDE.md; 5 of 6 areas found: commands, testing, code style, git workflow, boundaries |
| 4.3 | 1 | scan | Pass | Rule files that apply only to some paths | .claude/rules/*.md (17 files) |
| 4.4 | 1 | scan | Pass | Procedure files and agent definitions | .claude/skills/*/SKILL.md (2 files); .claude/agents/*.md (2 files) |
| 4.5 | 3 | scan | Pass | The model named in settings and in agent definitions | .claude/settings.json (env); .claude/agents/*.md (2 files) |
| 4.6 | 3 | scan | Pass | A hook on the tool that starts another agent (Claude Code PreToolUse on Agent; on other platforms, shown to a person) | .claude/settings.json: PreToolUse on Agent |
| 5.1 | 1 | survey answer | Pass | Your answer to the survey question "Does each agent work in its own checkout or environment?" | the answer was yes |
| 5.2 | 3 | scan | Pass | A hook before shell commands (Claude Code PreToolUse on Bash; Cursor beforeShellExecution; on other platforms, shown to a person) | .claude/settings.json: PreToolUse on Bash |
| 5.3 | 2 | shown to a person | Pass | The limit that stops a build or test run that hangs | the artifact was shown |
| 5.4 | 1 | survey answer | Pass | Your answer to the survey question "Is there a written limit on the size of 1 agent task?" | the answer was yes |
| 6.1 | 1 | scan | Pass | Command lines of the instruction files | CLAUDE.md; ~/.claude/CLAUDE.md |
| 6.2 | 2 | scan | Pass | A file in .git/hooks/ that is not a sample, or a workflow file in .github/workflows/ | .git/hooks/pre-commit; .git/hooks/pre-merge-commit |
| 6.3 | 3 | scan | Fail | A hook on the end of the agent's turn (Claude Code and Codex CLI Stop; Cursor stop; VS Code agentStop; on other platforms, shown to a person) | looked in: .claude/settings.json, .claude/settings.local.json, ~/.claude/settings.json; no hook registered on Stop |
| 6.4 | 4 | shown to a person | Pass | Test results kept as files for each build | the artifact was shown |
| 6.5 | 4 | shown to a person | Fail | For each function you name as crucial: a test set run at every release | there is no artifact |
| 6.6 | 2 | shown to a person | Pass | The gate that fails on new warnings in changed files | the artifact was shown |
| 7.1 | 1 | survey answer | Pass | Your answer to the survey question "Is every agent change reviewed before it merges by a reviewer other than its author: a person, or a model run as a separate reviewer?" | the answer was yes |
| 7.2 | 2 | shown to a person | Fail | A CODEOWNERS file; when none is found, branch protection shown to a person | looked in: .github/CODEOWNERS, CODEOWNERS, docs/CODEOWNERS; no CODEOWNERS file; branch protection is not on disk and is shown to a person; there is no artifact |
| 7.3 | 2 | shown to a person | Fail | An automated review that runs on every change before it merges and cannot be skipped | there is no artifact |
| 7.4 | 2 | shown to a person | Pass | The merge step that builds and checks after each merge and stops on failure | the artifact was shown |
| 7.5 | 1 | survey answer | Pass | Your answer to the survey question "Is it written who may start or assign an agent?" | the answer was yes |
| 7.6 | 3 | shown to a person | Pass | The written rule that says which changes need a person's approval before they merge or ship, by risk | the artifact was shown |
| 7.7 | 4 | shown to a person | Fail | The record of automated approvals, and the sample of them a person reviewed in the last quarter | there is no artifact |
| 8.1 | 1 | survey answer | Pass | Your answer to the survey question "Does a person confirm each release?" | the answer was yes |
| 8.2 | 2 | shown to a person | Pass | The release gate that refuses when a required item is missing | the artifact was shown |
| 8.3 | 4 | shown to a person | Pass | The record of each release's gate result | the artifact was shown |
| 9.1 | 3 | scan | Pass | The sandbox setting | ~/.claude/settings.json |
| 9.2 | 3 | scan | Pass | The setting that limits network access | ~/.claude/settings.json |
| 9.3 | 3 | scan | Pass | The permission rules in the agent's settings | .claude/settings.json; .claude/settings.local.json; ~/.claude/settings.json |
| 9.4 | 3 | scan | Pass | A hook that can deny a tool call (Claude Code and Codex CLI PreToolUse; Cursor preToolUse, beforeShellExecution, beforeMCPExecution, beforeReadFile; VS Code preToolUse; on other platforms, shown to a person) | .claude/settings.json: PreToolUse on Agent; .claude/settings.json: PreToolUse on Bash; .claude/settings.json: PreToolUse on an MCP tool; .claude/settings.json: PreToolUse on Write|Edit|NotebookEdit |
| 9.5 | 3 | scan | Fail | What happens when the hook itself fails | looked in: .claude/settings.json, .claude/settings.local.json, ~/.claude/settings.json; 0 of 4 hook entries have onFailure set to block |
| 9.6 | 2 | shown to a person | Pass | The secret scan that runs before code leaves the machine | the artifact was shown |
| 9.7 | 1 | scan | Pass | Lines of the instruction files that hold "never", "must not" or "do not" | CLAUDE.md; ~/.claude/CLAUDE.md; 35 boundary lines found |
| 9.8 | 4 | shown to a person | Pass | A list of the team's rules marked enforced, checked or written only | the artifact was shown |
| 10.1 | 1 | scan | Pass | How many files the agent's memory folder holds, never their contents | ~/.claude/projects/<project>/memory/; 63 files in the folder |
| 10.2 | 2 | shown to a person | Pass | The end-of-session step that refuses to finish with failing tests or unpushed work | the artifact was shown |
| 10.3 | 1 | survey answer | Pass | Your answer to the survey question "Is the start-of-session procedure written?" | the answer was yes |
| 10.4 | 1 | survey answer | Pass | Your answer to the survey question "Does each agent write down what it learned from its work, in a place the next agent reads?" | the answer was yes |
| 10.5 | 2 | shown to a person | Pass | The mechanism that refuses a change whose agent recorded no lesson entry | the artifact was shown |
| 10.6 | 2 | shown to a person | Fail | For the last 5 agent changes, the record that the agent read the carried-forward lessons before its work | there is no artifact |
| 11.1 | 4 | scan | Pass | The count and dates of the agent's session records, never their text, and only when you turn the count on | ~/.claude/projects/<project>/<session>.jsonl; 16 of 22 records are dated in the last 30 days |
| 11.2 | 4 | shown to a person | Fail | Merge rate and revert rate of agent changes against the team's own baseline | there is no artifact |
| 11.3 | 4 | shown to a person | Pass | Usage or cost recorded per agent or task | the artifact was shown |
| 11.4 | 4 | shown to a person | Fail | A quality trend: warnings, complexity or failures over time | there is no artifact |
| 11.5 | 1 | survey answer | Pass | Your answer to the survey question "Do you measure agent work by a named outcome?" | the answer was yes |
| 12.1 | 1 | survey answer | Unknown | Your answer to the survey question "Is there a written statement of which AI tools are used, on which data, and who approves?" | the question was not asked |
| 12.2 | 3 | shown to a person | Fail | The organization's policy file for the agent | looked in: /Library/Application Support/ClaudeCode/managed-settings.json, /Library/Application Support/ClaudeCode/managed-settings.d/*.json; no file at the documented locations; a policy delivered by a macOS configuration profile, the Windows registry or the vendor's server is not a file and is shown to an assessor; there is no artifact |
| 12.3 | 1 | shown to a person | Pass | Rules with stable identifiers and a record of why each exists | the artifact was shown |
| 12.4 | 1 | survey answer | Unknown | Your answer to the survey question "Is agent use covered by your organization's AI risk process?" | the question was not asked |
| 13.1 | 1 | survey answer | Pass | Your answer to the survey question "Is it written which model and which reasoning effort each kind of task uses?" | the answer was yes |
| 13.2 | 1 | scan | Pass | The number of lines in each instruction file, never the lines | CLAUDE.md; ~/.claude/CLAUDE.md; the longest instruction file has 194 lines |
| 13.3 | 3 | scan | Pass | The setting for compacting the conversation | .claude/settings.json |
| 13.4 | 3 | scan | Pass | The reasoning-effort setting | ~/.claude/settings.json |
| 13.5 | 3 | scan | Pass | The model set for agents that another agent starts | .claude/settings.json; .claude/agents/*.md (2 files) |
| 13.6 | 3 | scan | Pass | The prompt-cache settings | .claude/settings.json |
| 13.7 | 4 | shown to a person | Pass | A report that separates input tokens read from the cache from input tokens sent uncached | the artifact was shown |
| 13.8 | 2 | shown to a person | Pass | A mechanism that shortens long command output before the agent reads it | the artifact was shown |
| 13.9 | 1 | survey answer | Pass | Your answer to the survey question "Is it written that long or verbose work goes to its own agent, and that agents wait without polling?" | the answer was yes |
| 13.10 | 3 | shown to a person | Pass | A spend or usage limit set at the provider, for the workspace or for each person | the artifact was shown |
| 13.11 | 4 | shown to a person | Pass | A review of tokens or cost per model over time, and 1 change made because of it | the artifact was shown |
| 13.12 | 4 | scan | Fail | The usage-telemetry setting | looked in: ~/.claude/settings.json; the setting is not set |
| 14.1 | 3 | scan | Pass | A hook that runs before the agent writes or edits a file (Claude Code PreToolUse on Write or Edit) | .claude/settings.json: PreToolUse on Write|Edit|NotebookEdit |
| 14.2 | 1 | scan | Pass | The MCP server declarations and the permission rules | .claude/settings.json |
| 14.3 | 3 | scan | Pass | The tool limits in agent definitions | .claude/agents/*.md (1 file) |
| 14.4 | 1 | scan | Pass | Worktree folders and the worktree setup file | .claude/worktrees/* (2 folders) |
| 14.5 | 2 | shown to a person | Pass | The command that shows the state of each background or unattended run | the artifact was shown |
| 14.6 | 3 | shown to a person | Pass | The channel through which an agent runs a command on another machine | the artifact was shown |
| 14.7 | 2 | shown to a person | Pass | The tests of the team's own hooks and guard scripts | the artifact was shown |
| 15.1 | 3 | shown to a person | Pass | The script every delete goes through, and the hook that refuses a delete made any other way | the artifact was shown |
| 15.2 | 3 | shown to a person | Pass | The hook that refuses a build or test command run outside the team's wrapper script | the artifact was shown |
| 15.3 | 2 | shown to a person | Pass | The check that keeps credentials and real messages out of test fixtures | the artifact was shown |
| 15.4 | 1 | survey answer | Pass | Your answer to the survey question "Is it written that a new integration or tool starts with no access until it is listed?" | the answer was yes |
| 15.5 | 2 | shown to a person | Pass | What stops a test from reaching a live account unless a person turns that on | the artifact was shown |
| 15.6 | 2 | shown to a person | Pass | The commit check that refuses personal identifiers and authorship marks | the artifact was shown |
| 15.7 | 2 | shown to a person | Pass | The set of injection attempts run against a model before it is given a channel, and the step that runs it | the artifact was shown |
| 15.8 | 1 | survey answer | Pass | Your answer to the survey question "Is it written that a change touches only what its own work item needs?" | the answer was yes |
| 15.9 | 2 | shown to a person | Pass | For 1 guard: its single switch, and the test of the off state | the artifact was shown |
| 15.10 | 3 | shown to a person | Pass | The script that removes an agent's workspace and refuses when its work is not merged | the artifact was shown |
| 15.11 | 3 | scan | Pass | The setting that lets a command run outside the sandbox | ~/.claude/settings.json |
| 16.1 | 1 | survey answer | Pass | Your answer to the survey question "Is there a written record of how each model you use has failed in your own work?" | the answer was yes |
| 16.2 | 1 | survey answer | Pass | Your answer to the survey question "Does every brief to an agent state what done means?" | the answer was yes |
| 16.3 | 2 | shown to a person | Pass | For the last 2 agent reports: who other than the agent checked the claims, and what they checked | the artifact was shown |
| 16.4 | 2 | scan | Fail | A hook on a denied tool call (Claude Code PermissionDenied; Cursor postToolUseFailure; on other platforms, shown to a person) | looked in: .claude/settings.json, .claude/settings.local.json, ~/.claude/settings.json; no hook registered on PermissionDenied |
| 16.5 | 2 | shown to a person | Pass | The log of refused calls, each with the model that made the call | the artifact was shown |
| 16.6 | 3 | shown to a person | Fail | The mechanism that refuses the end of a turn while tasks are open, and 1 refusal it made | there is no artifact |
| 16.7 | 3 | scan | Fail | A hook on the end of a started agent's work or on a task being marked complete (Claude Code SubagentStop, TaskCompleted; Codex CLI SubagentStop; Cursor subagentStop; VS Code subagentStop; on other platforms, shown to a person) | looked in: .claude/settings.json, .claude/settings.local.json, ~/.claude/settings.json; no hook registered on SubagentStop, TaskCompleted |
| 16.8 | 4 | shown to a person | Fail | Counts per model for 2 periods (rule breaks, early stops, false "done" reports) and the schedule on which they are read | there is no artifact |
| 16.9 | 4 | shown to a person | Fail | The measurement made before the last new model version was adopted, and the line of the model policy it changed or confirmed | there is no artifact |
| Level | Name | What it takes |
|---|---|---|
| AEI Level 0 | Unrecorded | No evidence for the practice |
| AEI Level 1 | Written | The practice is written where the agent or the team reads it |
| AEI Level 2 | Checked | A mechanism checks the result at commit, merge or release, and refuses |
| AEI Level 3 | Enforced | A mechanism refuses the action before it happens |
| AEI Level 4 | Measured | Records of what happened are kept and read |