duediligence 0.3.1 → 0.4.0
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/package.json +1 -1
- package/skills/duediligence/SKILL.md +15 -7
package/package.json
CHANGED
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
{
|
|
2
2
|
"name": "duediligence",
|
|
3
|
-
"version": "0.
|
|
3
|
+
"version": "0.4.0",
|
|
4
4
|
"description": "Due diligence for AI coding agents: independent, hidden test suites written from the spec. The agent works against a public suite; a hidden suite it never sees judges the result.",
|
|
5
5
|
"keywords": [
|
|
6
6
|
"due-diligence",
|
|
@@ -10,7 +10,7 @@ description: >-
|
|
|
10
10
|
spec, or wants to stop trusting an agent's own green test run.
|
|
11
11
|
license: MIT
|
|
12
12
|
compatibility: >-
|
|
13
|
-
Node 20.19 or later. Runs the duediligence@0.
|
|
13
|
+
Node 20.19 or later. Runs the duediligence@0.4.0 CLI from npm, which starts
|
|
14
14
|
Claude Code, Codex or Anthropic API sessions and installs its test toolchain
|
|
15
15
|
under ~/.duediligence/ once (network).
|
|
16
16
|
---
|
|
@@ -20,10 +20,10 @@ compatibility: >-
|
|
|
20
20
|
From one confirmed spec, generate a **public** suite the implementing agent
|
|
21
21
|
works against and a **hidden** suite that judges the result.
|
|
22
22
|
|
|
23
|
-
`duediligence <command>` below is the CLI, `duediligence@0.
|
|
23
|
+
`duediligence <command>` below is the CLI, `duediligence@0.4.0`. When
|
|
24
24
|
`duediligence --version` prints another version or nothing, ask the person
|
|
25
|
-
before installing it: `npm i -g duediligence@0.
|
|
26
|
-
`npx duediligence@0.
|
|
25
|
+
before installing it: `npm i -g duediligence@0.4.0`, or put
|
|
26
|
+
`npx duediligence@0.4.0` in place of `duediligence` in each command (in a clone
|
|
27
27
|
of the repository, `node bin/cli.mjs`). Run it from any folder. `specs/` and
|
|
28
28
|
`runs/` below are in the home, `~/.duediligence/` (`DUEDILIGENCE_HOME`);
|
|
29
29
|
`prompts/` and `template*/` are in the CLI's package. `duediligence where`
|
|
@@ -88,10 +88,18 @@ bundles the server; `/mcp` signs in), they work on the person's workspace on
|
|
|
88
88
|
the Lunox platform. Reads: `list_workspaces`, `list_bounties`, `get_bounty`,
|
|
89
89
|
`get_proposal`, `get_run`, `get_usage`. Writes, only when the person asks for
|
|
90
90
|
that one: `post_bounty`, `import_github_issue`, `import_jira_issue` (a new
|
|
91
|
-
bounty) and `size_bounty` (price it; this starts a paid model run).
|
|
91
|
+
bounty) and `size_bounty` (price it; this starts a paid model run).
|
|
92
|
+
`get_decisions` reads what a priced spec decided where the ticket was silent
|
|
93
|
+
(its `Decided:` items) and whether a person confirmed them; `confirm_link`
|
|
94
|
+
gives the page where the person confirms them. Hand them that link: decision
|
|
92
95
|
confirmations, approving a price and publishing happen in the platform's web
|
|
93
|
-
app
|
|
94
|
-
|
|
96
|
+
app, and you never act for them. Once the decisions are confirmed and the
|
|
97
|
+
person asks: `start_sandbox` and `prepare` (build a sandbox version; the
|
|
98
|
+
platform writes hidden tests from the confirmed spec and calibrates them
|
|
99
|
+
against a reference, as `prepare` does locally), then `get_handout` for the
|
|
100
|
+
public files and the hidden suite's commitment hash. You never see the hidden
|
|
101
|
+
tests in hosted mode: they stay on the platform. Judging a submission is not
|
|
102
|
+
hosted yet; use local mode for that. Never ask
|
|
95
103
|
for a token in the chat: the server signs in through `/mcp`, the CLI through
|
|
96
104
|
`duediligence login` in the person's own terminal. Local mode, the steps below,
|
|
97
105
|
is unchanged and the default, and needs no sign-in.
|