@jjanczur/tyran 0.1.39 → 0.1.42
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/.claude-plugin/marketplace.json +1 -1
- package/.claude-plugin/plugin.json +1 -1
- package/CHANGELOG.md +225 -0
- package/README.md +6 -3
- package/agents/implementer.md +6 -3
- package/agents/retro.md +10 -7
- package/agents/reviewer.md +4 -2
- package/agents/scout.md +3 -2
- package/agents/verifier.md +3 -2
- package/bin/tyran.mjs +1 -1
- package/package.json +4 -1
- package/scripts/board-html.mjs +264 -19
- package/scripts/board.mjs +134 -6
- package/scripts/doctor.mjs +116 -4
- package/scripts/journal.mjs +17 -1
- package/scripts/knowledge.mjs +219 -20
- package/scripts/project.mjs +9 -1
- package/scripts/schema.mjs +37 -2
- package/skills/browser-check/SKILL.md +1 -1
- package/skills/code-review/SKILL.md +1 -1
- package/skills/deslop/SKILL.md +1 -1
- package/skills/fidelity-gate/SKILL.md +1 -1
- package/skills/pr-feedback/SKILL.md +1 -1
- package/skills/prompt-tuning/SKILL.md +1 -1
- package/skills/retro/SKILL.md +27 -1
- package/skills/root-cause/SKILL.md +1 -1
- package/skills/run/SKILL.md +5 -3
- package/skills/skill-writing/SKILL.md +1 -1
- package/templates/knowledge.yaml +5 -0
|
@@ -1,5 +1,5 @@
|
|
|
1
1
|
---
|
|
2
|
-
description: Work a reviewer's comments on a pull request to the end
|
|
2
|
+
description: Work a reviewer's comments on a pull request to the end - all three comment surfaces enumerated before any is triaged, every comment fixed, declined with a reason, or ticketed, push before you reply. Use when a PR comes back with review feedback or a red check.
|
|
3
3
|
---
|
|
4
4
|
|
|
5
5
|
# PR feedback — closing the loop with a human reviewer
|
|
@@ -1,5 +1,5 @@
|
|
|
1
1
|
---
|
|
2
|
-
description:
|
|
2
|
+
description: Tune a prompt, or anything whose quality is measured by non-deterministic model output, without chasing noise - a noise baseline before the first edit, medians over repeated runs, enforcement AFTER generation rather than in the wording. Use when iterating on prompts or model-judged output.
|
|
3
3
|
---
|
|
4
4
|
|
|
5
5
|
# Prompt tuning — working with non-deterministic output
|
package/skills/retro/SKILL.md
CHANGED
|
@@ -1,5 +1,5 @@
|
|
|
1
1
|
---
|
|
2
|
-
description: Close an initiative by learning from it
|
|
2
|
+
description: Close an initiative by learning from it - spawns the retrospective agent over the ledger, the notes and the agents' reports, records what changed and what was rejected, and writes durable repo facts into .tyran/knowledge/. Runs at the end of an initiative; also available as /tyran:retro.
|
|
3
3
|
---
|
|
4
4
|
|
|
5
5
|
# Retro
|
|
@@ -66,6 +66,32 @@ here, at close, in the one place licensed to write these files. Keep entries
|
|
|
66
66
|
under the size `doctor --state` warns about (`knowledge-entry-oversized`,
|
|
67
67
|
4 000 codepoints); an oversized entry crowds every brief it matches.
|
|
68
68
|
|
|
69
|
+
**Then consolidate — by APPENDING, never by editing.** Read the audit:
|
|
70
|
+
|
|
71
|
+
```bash
|
|
72
|
+
node "${CLAUDE_PLUGIN_ROOT}/scripts/knowledge.mjs" audit --dir .tyran/knowledge
|
|
73
|
+
```
|
|
74
|
+
|
|
75
|
+
It names the retirement candidates the counters have written off, and — if it
|
|
76
|
+
says no entry carries counter evidence — that the fold above is not actually
|
|
77
|
+
happening, which is the finding to act on first.
|
|
78
|
+
|
|
79
|
+
Deciding that two entries say the same thing is YOUR judgement; no script
|
|
80
|
+
scores it. Recording that decision is mechanical: write a **new** file,
|
|
81
|
+
`.tyran/knowledge/<slug>-merged.yaml`, whose entry states the fact once and
|
|
82
|
+
carries `supersedes:` naming every id it replaces, plus the union of their
|
|
83
|
+
`provenance`. **Do not touch the files holding the originals.** They keep
|
|
84
|
+
their bytes and the counters they earned over months, and stop reaching briefs
|
|
85
|
+
because `supersedes` retires them at read time. An entry the counters have
|
|
86
|
+
written off with nothing to replace it is a plain deletion — git holds the
|
|
87
|
+
bytes, and the counters are the argument.
|
|
88
|
+
|
|
89
|
+
Re-run the audit. Superseded should be up, `live` down, reachability up, and
|
|
90
|
+
**dangling zero** — a `supersedes` naming an id that is not in the store means
|
|
91
|
+
the retirement silently did not happen and the merge only made the store
|
|
92
|
+
bigger. Undo, if the merge was wrong, is deleting that one file: every entry
|
|
93
|
+
it retired comes back whole.
|
|
94
|
+
|
|
69
95
|
Write what a competent newcomer would get wrong, not what the code already
|
|
70
96
|
says. "The test suite takes 9 minutes, so do not run it per file" is worth an
|
|
71
97
|
entry. "This is a Next.js app" is not: the next agent can see that.
|
|
@@ -1,5 +1,5 @@
|
|
|
1
1
|
---
|
|
2
|
-
description: Find the mechanism behind a failure instead of patching its symptom
|
|
2
|
+
description: Find the mechanism behind a failure instead of patching its symptom - reproduce first, one variable per experiment with the prediction written before the run, exit by naming the mechanism and pinning it with a failing test. Use for a bug, an unexplained red test, or a failure that will not reproduce.
|
|
3
3
|
---
|
|
4
4
|
|
|
5
5
|
# Root cause — working a failure down to its mechanism
|
package/skills/run/SKILL.md
CHANGED
|
@@ -1,5 +1,5 @@
|
|
|
1
1
|
---
|
|
2
|
-
description: Conduct a task end to end, from a one-line fix to a multi-day programme
|
|
2
|
+
description: Conduct a task end to end, from a one-line fix to a multi-day programme - a short interview, then autonomous execution to the finish. Classifies the task S/M/L/XL, configures itself to the repo, routes each role to a cost tier, drives the agent team from the kanban board, and stops only at genuine decisions. Use for /tyran:run or when asked to carry something "all the way" or "with a team of agents".
|
|
3
3
|
---
|
|
4
4
|
|
|
5
5
|
# Tyran — the conductor
|
|
@@ -256,8 +256,10 @@ open one.
|
|
|
256
256
|
the next report, review or merge clears it.
|
|
257
257
|
- **Scout `DURABLE:` findings, you journal** as `finding` events — the
|
|
258
258
|
required keys are `area` + `claim` (the area is the path or subsystem the
|
|
259
|
-
scout's proof points at), plus that `proof` —
|
|
260
|
-
|
|
259
|
+
scout's proof points at), plus that `proof` — and, when the scout proved
|
|
260
|
+
it by running something, the `command` and its `exit_code`. Scouts write
|
|
261
|
+
nothing themselves; implementers, reviewers and verifiers append their
|
|
262
|
+
own. An operator
|
|
261
263
|
CORRECTION — a premise you had wrong, a step they had to undo — is a
|
|
262
264
|
`finding` too, with `area: process`: it is the densest mistake signal
|
|
263
265
|
there is, and the retrospective is where it becomes a `MISTAKES.md`
|
|
@@ -1,5 +1,5 @@
|
|
|
1
1
|
---
|
|
2
|
-
description: How a skill earns its place here
|
|
2
|
+
description: How a skill earns its place here and how it is written so it fires - a three-question admission test, one protocol per skill, the description priced against the CI budget, and an activation test from a cold session. Use when the retrospective proposes a new skill or rewrites one.
|
|
3
3
|
---
|
|
4
4
|
|
|
5
5
|
# Writing a skill for this plugin
|
package/templates/knowledge.yaml
CHANGED
|
@@ -6,6 +6,11 @@
|
|
|
6
6
|
#
|
|
7
7
|
# kind: fact | convention | gotcha | command | decision
|
|
8
8
|
#
|
|
9
|
+
# To merge two entries that say one thing, APPEND a third that states it once
|
|
10
|
+
# and carries `supersedes: [K-3, K-7]`. The named entries drop out of every
|
|
11
|
+
# brief while keeping their bytes and counters, so deleting the new file
|
|
12
|
+
# brings them back. Never edit or delete the originals to merge them.
|
|
13
|
+
#
|
|
9
14
|
# Validate: node scripts/schema.mjs validate knowledge .tyran/knowledge/*.yaml
|
|
10
15
|
|
|
11
16
|
entries:
|