pi-braid 0.43.22 → 0.43.23
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/CHANGELOG.md +42 -0
- package/build-info.json +4 -4
- package/package.json +1 -1
- package/skills/agent-testing/SKILL.md +7 -3
- package/skills/braid-reference/SKILL.md +6 -2
- package/skills/code-refactoring/SKILL.md +6 -0
- package/skills/code-search/SKILL.md +4 -0
- package/skills/documentation/SKILL.md +5 -0
- package/skills/dox/SKILL.md +5 -0
- package/skills/healthcheck/SKILL.md +7 -2
- package/skills/meta-cognition/SKILL.md +7 -1
- package/skills/meta-improvement/SKILL.md +6 -3
- package/skills/misdirection-audit/SKILL.md +5 -2
- package/skills/pi-reference/SKILL.md +6 -2
- package/skills/placement-discipline/SKILL.md +6 -2
- package/skills/probe-evaluation/SKILL.md +6 -2
- package/skills/release-pipeline/SKILL.md +6 -2
- package/skills/research/SKILL.md +6 -0
- package/skills/skill-usage/SKILL.md +6 -0
- package/skills/software-engineer/SKILL.md +5 -2
- package/skills/subagent-orchestration/SKILL.md +7 -2
- package/skills/subagent-verification/SKILL.md +6 -1
- package/skills/tone-quality/SKILL.md +6 -3
- package/skills/voice-replies/SKILL.md +7 -2
package/CHANGELOG.md
CHANGED
|
@@ -53,6 +53,48 @@ record; the tag scheme above applies unchanged.
|
|
|
53
53
|
|
|
54
54
|
---
|
|
55
55
|
|
|
56
|
+
## Contrib — 0.1.4 (2026-11-04: the cold rehearsal measures what it boots)
|
|
57
|
+
|
|
58
|
+
- **The spawn is timed where it happens.** It sat outside the timed check, so a
|
|
59
|
+
step that installs 211 packages and calls a model reported 0ms, a duration no
|
|
60
|
+
reader should believe and a cost the report then hid.
|
|
61
|
+
- **The child is the managed CLI, not whatever launched the probe.** The path came
|
|
62
|
+
from `process.argv[1]`, which is the pi CLI only when pi launched the process; a
|
|
63
|
+
rehearsal started from a script re-executed its own entry point and reported the
|
|
64
|
+
resulting module error as a probe failure.
|
|
65
|
+
- **Credential resolution state is copied with the config.** Without the provider
|
|
66
|
+
key-ring and route files a fresh home answers "No API key found for the selected
|
|
67
|
+
model" and the probe never runs, which is also the true cause of a silent run
|
|
68
|
+
this instrument once reported as unexplained.
|
|
69
|
+
- **A silent attempt buys exactly one retry**, because one attempt's silence does
|
|
70
|
+
not say why: the retry named the missing credential outright.
|
|
71
|
+
|
|
72
|
+
## Skills — 0.2.5 (2026-11-04: the contract reads both skill trees)
|
|
73
|
+
|
|
74
|
+
- **One implementation, two populations.** `skills-contract` now scans the pack and
|
|
75
|
+
the core tree, prints both counts, and fails closed on a tree that yields no
|
|
76
|
+
`SKILL.md`. Phrase uniqueness is checked across both: one phrase names one skill.
|
|
77
|
+
- **The tool allowlist stays pack-scoped**, because a core skill legitimately names
|
|
78
|
+
tools the pack cannot reach. A foreign tool named by a pack skill still fails.
|
|
79
|
+
- **The gate is in the chain**, so the core tree can no longer drift out of the
|
|
80
|
+
contract; a new fixture covers both trees, the scoping and the absent-versus-empty
|
|
81
|
+
populations.
|
|
82
|
+
|
|
83
|
+
## Core — 0.43.23 (2026-11-04: 21 core skills declare their phrases where the matcher reads them)
|
|
84
|
+
|
|
85
|
+
- **The core skill tree follows the contract it documents.** Of 23 skills, 21 moved
|
|
86
|
+
from a legacy `trigger:` key, a body `## trigger:` heading the validator refuses,
|
|
87
|
+
or no phrase at all, into a frontmatter `triggers:` list. Seven of them had no
|
|
88
|
+
phrase the matcher could read.
|
|
89
|
+
- **Three descriptions lose a `Triggers:` tail** the reader silently drops past 200
|
|
90
|
+
characters, and five are rewritten to name the situation they serve. No body text
|
|
91
|
+
changed: the migration moved declarations and nothing else.
|
|
92
|
+
- **Reach is not proven by a green gate.** The contract counts, lengths and
|
|
93
|
+
uniqueness; it cannot test whether a phrase appears in anything a person would
|
|
94
|
+
type. Measured on the trigger leg alone, five of eight natural prompts in a
|
|
95
|
+
third-party sample matched no phrase, so the next round belongs to adoption data
|
|
96
|
+
rather than another guess.
|
|
97
|
+
|
|
56
98
|
## Contrib — 0.1.3 (2026-11-03: the health check rehearses a machine that has no state)
|
|
57
99
|
|
|
58
100
|
- **A `coldstart` category answers the question the other categories cannot.** Every existing probe
|
package/build-info.json
CHANGED
|
@@ -1,9 +1,9 @@
|
|
|
1
1
|
{
|
|
2
2
|
"name": "pi-braid",
|
|
3
|
-
"version": "0.43.
|
|
4
|
-
"sha": "
|
|
5
|
-
"shaShort": "
|
|
3
|
+
"version": "0.43.23",
|
|
4
|
+
"sha": "e63d33701040689c94aab6c50ebb067d5ef94df3",
|
|
5
|
+
"shaShort": "e63d3370",
|
|
6
6
|
"branch": "detached",
|
|
7
7
|
"dirty": false,
|
|
8
|
-
"packedAt": "2026-11-
|
|
8
|
+
"packedAt": "2026-11-03T23:20:23.106Z"
|
|
9
9
|
}
|
package/package.json
CHANGED
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
{
|
|
2
2
|
"name": "pi-braid",
|
|
3
|
-
"version": "0.43.
|
|
3
|
+
"version": "0.43.23",
|
|
4
4
|
"description": "🥨 Braid - Metacognitive extension suite for pi coding agent: Adds memory, plan tracking, structured cognitive signals and completion checks on top of pi core loop, so long sessions keep their context and stop claiming unfinished work.",
|
|
5
5
|
"type": "module",
|
|
6
6
|
"main": "./dist/index.js",
|
|
@@ -1,12 +1,16 @@
|
|
|
1
1
|
---
|
|
2
2
|
name: agent-testing
|
|
3
|
-
description: "Run and evaluate coding agents programmatically: drive pi in RPC mode (JSON events over stdin/stdout), parse agent event streams, run multi-trial benchmarks, and verify tool execution results. Use when building a test harness for an agent, benchmarking or comparing agent configurations and models, asserting on the tool calls an agent made, or extracting structured results from agent runs.
|
|
3
|
+
description: "Run and evaluate coding agents programmatically: drive pi in RPC mode (JSON events over stdin/stdout), parse agent event streams, run multi-trial benchmarks, and verify tool execution results. Use when building a test harness for an agent, benchmarking or comparing agent configurations and models, asserting on the tool calls an agent made, or extracting structured results from agent runs."
|
|
4
|
+
triggers:
|
|
5
|
+
- "rpc mode"
|
|
6
|
+
- "agent benchmark"
|
|
7
|
+
- "test harness"
|
|
8
|
+
- "multi-trial"
|
|
9
|
+
- "tool execution verification"
|
|
4
10
|
validated: 2026-09-09
|
|
5
11
|
verdict: KEEP
|
|
6
12
|
---
|
|
7
13
|
|
|
8
|
-
## trigger: "benchmarking, multi-trial comparisons, tool execution verification, evaluating agent configurations, building test suites"
|
|
9
|
-
|
|
10
14
|
# Agent Testing
|
|
11
15
|
|
|
12
16
|
Test agent behavior programmatically. pi runs in RPC mode as a subprocess - you send JSON events, receive structured events back.
|
|
@@ -1,12 +1,16 @@
|
|
|
1
1
|
---
|
|
2
2
|
name: braid-reference
|
|
3
3
|
description: "Find braid's own documentation: context policy and compaction, timers and wakes, workers and the orchestrator registry, the Telegram channel, prompt and discipline rules, memory stores, the skills catalog, release and install paths, health maintenance, extensions and the polyfill. Use when a question is about braid internals, before editing braid, or when a braid behavior needs its owning doc named instead of guessed."
|
|
4
|
+
triggers:
|
|
5
|
+
- "braid docs"
|
|
6
|
+
- "braid documentation"
|
|
7
|
+
- "braid internals"
|
|
8
|
+
- "braid compaction"
|
|
9
|
+
- "braid timers"
|
|
4
10
|
validated: 2026-10-18
|
|
5
11
|
verdict: KEEP
|
|
6
12
|
---
|
|
7
13
|
|
|
8
|
-
## trigger: "braid docs, pi-braid documentation, braid context policy, braid compaction bands, braid timers and wakes, braid workers, orchestrator registry, braid telegram channel, braid prompt rules, braid memory stores, braid skills catalog, braid release, braid install paths, braid health maintenance, braid extension, polyfill"
|
|
9
|
-
|
|
10
14
|
# Braid Reference
|
|
11
15
|
|
|
12
16
|
Where braid's own knowledge lives, and which file owns which question. Two readers use this: an agent answering a question about braid, and a contributor about to edit braid. For the second, the owning `AGENTS.md` chain is a binding contract - read it before the first edit (the `dox` skill defines that tree).
|
|
@@ -1,6 +1,12 @@
|
|
|
1
1
|
---
|
|
2
2
|
name: code-refactoring
|
|
3
3
|
description: "Use when restructure/move/rename files or symbols across a TypeScript codebase. Covers ast-grep import path rewrites, file moves, symbol renames, and multi-file batch updates. Always use ast-grep for structural changes over sed/regex."
|
|
4
|
+
triggers:
|
|
5
|
+
- "refactor"
|
|
6
|
+
- "codemod"
|
|
7
|
+
- "ast-grep"
|
|
8
|
+
- "rename a symbol"
|
|
9
|
+
- "update imports"
|
|
4
10
|
validated: 2026-08-16
|
|
5
11
|
verdict: IMPROVE
|
|
6
12
|
---
|
|
@@ -3,6 +3,10 @@ name: code-search
|
|
|
3
3
|
description: "Use when searching code, refactoring, running codemods, or choosing between search tools. Covers ast-grep vs ripgrep selection and usage patterns."
|
|
4
4
|
status: merged
|
|
5
5
|
successor: code-refactoring
|
|
6
|
+
triggers:
|
|
7
|
+
- "code search"
|
|
8
|
+
- "search tool"
|
|
9
|
+
- "ripgrep vs ast-grep"
|
|
6
10
|
validated: 2026-08-16
|
|
7
11
|
verdict: MOVE
|
|
8
12
|
---
|
|
@@ -1,6 +1,11 @@
|
|
|
1
1
|
---
|
|
2
2
|
name: documentation
|
|
3
3
|
description: "Analyze a repository and generate architecture documentation, README.md, or a full documentation ecosystem. Use when documenting a codebase, summarizing architecture, or creating project docs."
|
|
4
|
+
triggers:
|
|
5
|
+
- "readme"
|
|
6
|
+
- "architecture docs"
|
|
7
|
+
- "codebase summary"
|
|
8
|
+
- "document the codebase"
|
|
4
9
|
validated: 2026-08-16
|
|
5
10
|
verdict: IMPROVE
|
|
6
11
|
---
|
package/skills/dox/SKILL.md
CHANGED
|
@@ -1,6 +1,11 @@
|
|
|
1
1
|
---
|
|
2
2
|
name: dox
|
|
3
3
|
description: "Maintain AGENTS.md hierarchy — agent-maintained living documentation tree with pre-edit walk and post-change update. Use when project guidelines drift, AGENTS.md files are missing or stale, or you need to set up or repair the documentation hierarchy. Use before editing files in unfamiliar or undocumented areas."
|
|
4
|
+
triggers:
|
|
5
|
+
- "agents.md"
|
|
6
|
+
- "agents.md hierarchy"
|
|
7
|
+
- "stale agents.md"
|
|
8
|
+
- "documentation hierarchy"
|
|
4
9
|
validated: 2026-08-16
|
|
5
10
|
verdict: KEEP
|
|
6
11
|
---
|
|
@@ -1,7 +1,12 @@
|
|
|
1
1
|
---
|
|
2
2
|
name: healthcheck
|
|
3
|
-
description: "Run a structured self-diagnostic of the agent environment: verify read/write/edit/bash/grep/find/ls, cognitive and memory tools, extension and skill availability, and ~/.pi writability; interpret pass/warn/fail results and clean up stale agent processes. Use when a tool misbehaves or fails, right after setup or a config change, before starting work in a fresh environment, or when many agent instances accumulated and the machine slows down.
|
|
4
|
-
|
|
3
|
+
description: "Run a structured self-diagnostic of the agent environment: verify read/write/edit/bash/grep/find/ls, cognitive and memory tools, extension and skill availability, and ~/.pi writability; interpret pass/warn/fail results and clean up stale agent processes. Use when a tool misbehaves or fails, right after setup or a config change, before starting work in a fresh environment, or when many agent instances accumulated and the machine slows down."
|
|
4
|
+
triggers:
|
|
5
|
+
- "healthcheck"
|
|
6
|
+
- "tool diagnostics"
|
|
7
|
+
- "stale processes"
|
|
8
|
+
- "post-setup"
|
|
9
|
+
- "environment readiness"
|
|
5
10
|
validated: 2026-11-03
|
|
6
11
|
verdict: IMPROVE
|
|
7
12
|
---
|
|
@@ -1,6 +1,12 @@
|
|
|
1
1
|
---
|
|
2
2
|
name: meta-cognition
|
|
3
|
-
description: "
|
|
3
|
+
description: "Use when an enforcement signal arrives or a transition is reached - between tasks, after tool calls, on signals. Signal-to-action mappings for enforcement whisper families (verification judge, tool loop, error burst, context advisory, unpersisted learnings, signal density warnings), then close the loop: observe, evaluate, adapt, persist."
|
|
4
|
+
triggers:
|
|
5
|
+
- "verification judge"
|
|
6
|
+
- "tool loop"
|
|
7
|
+
- "error burst"
|
|
8
|
+
- "context advisory"
|
|
9
|
+
- "unpersisted learning"
|
|
4
10
|
validated: 2026-08-16
|
|
5
11
|
verdict: KEEP
|
|
6
12
|
---
|
|
@@ -1,12 +1,15 @@
|
|
|
1
1
|
---
|
|
2
2
|
name: meta-improvement
|
|
3
|
-
description: "
|
|
3
|
+
description: "Use when a skill change has to be measured rather than assumed: run self-improving experiments and measure skill performance. Covers archive-based evolution with staged evaluation, and testing improvements against baselines. Not for authoring a skill; `skill-authoring` owns the file, its frontmatter and its trigger phrases."
|
|
4
|
+
triggers:
|
|
5
|
+
- "self-improvement experiment"
|
|
6
|
+
- "archive-based evolution"
|
|
7
|
+
- "staged evaluation"
|
|
8
|
+
- "skill performance"
|
|
4
9
|
validated: 2026-08-16
|
|
5
10
|
verdict: KEEP
|
|
6
11
|
---
|
|
7
12
|
|
|
8
|
-
## trigger: "self-improvement experiments, skill performance measurement, archive-based evolution, staged evaluation, testing improvements against baselines"
|
|
9
|
-
|
|
10
13
|
# Meta-Improvement
|
|
11
14
|
|
|
12
15
|
Run self-improving experiments against an archive that keeps every variant, not just the best.
|
|
@@ -1,12 +1,15 @@
|
|
|
1
1
|
---
|
|
2
2
|
name: misdirection-audit
|
|
3
3
|
description: "Periodic self-audit for misdirected work: run the placement checker, grep generic layers for role/path policy, byte-diff assembled prompts between builds, then classify findings (wrong-layer fix, fake-generic presentation, enforcement gap, unmeasured innovation, doc drift, and more) into one ledger row per instance, where a recurring class earns a gate. Use at wave boundaries, before a release, or when a task asks for a misdirection, placement, or hygiene audit."
|
|
4
|
+
triggers:
|
|
5
|
+
- "misdirection audit"
|
|
6
|
+
- "placement audit"
|
|
7
|
+
- "hygiene sweep"
|
|
8
|
+
- "self-audit"
|
|
4
9
|
validated: 2026-09-15
|
|
5
10
|
verdict: UNVALIDATED
|
|
6
11
|
---
|
|
7
12
|
|
|
8
|
-
## trigger: "misdirection audit, placement audit, hygiene sweep, wave boundary audit, audit recent output, periodic self-audit"
|
|
9
|
-
|
|
10
13
|
# Misdirection Audit
|
|
11
14
|
|
|
12
15
|
A periodic audit of the project's own recent output for work that went in the wrong direction: fixes in the wrong layer, rules that exist only as prose, metrics that cannot test what they alarm on. Output: classified instances, one ledger row each, and promotion decisions. The audit is read-only on code; it records, it does not fix.
|
|
@@ -1,12 +1,16 @@
|
|
|
1
1
|
---
|
|
2
2
|
name: pi-reference
|
|
3
3
|
description: "Use when asking about pi capabilities, configuration, customization, or extending pi. Covers SDK, extensions, skills, themes, TUI, keybindings, models, providers, prompts, RPC mode, session management, settings, CLI, and internal architecture. Also use when creating custom tools or building self-improving extensions."
|
|
4
|
+
triggers:
|
|
5
|
+
- "pi capabilities"
|
|
6
|
+
- "extension development"
|
|
7
|
+
- "extending pi"
|
|
8
|
+
- "pi docs"
|
|
9
|
+
- "pi settings"
|
|
4
10
|
validated: 2026-10-08
|
|
5
11
|
verdict: KEEP
|
|
6
12
|
---
|
|
7
13
|
|
|
8
|
-
## trigger: "pi capabilities, configuration, customization, extending pi, SDK usage, extension development, themes, keybindings, models, providers, prompts, RPC mode, session management, settings, CLI"
|
|
9
|
-
|
|
10
14
|
# Pi Reference
|
|
11
15
|
|
|
12
16
|
Complete reference for pi's architecture, configuration, and internals.
|
|
@@ -1,12 +1,16 @@
|
|
|
1
1
|
---
|
|
2
2
|
name: placement-discipline
|
|
3
3
|
description: "The operational workflow around placement discipline: name the layer before building, keep domain facts out of shared layers (a fix lives where the invariant lives), carry the required ## Placement section in impl/audit/verify reports, append repairs to the placement ledger, and tamper-test the checker before trusting a green run. Use when planning or reviewing any change that touches a shared or generic layer, when writing a report, or when validating a worker's result."
|
|
4
|
+
triggers:
|
|
5
|
+
- "wrong layer"
|
|
6
|
+
- "generic layer"
|
|
7
|
+
- "placement ledger"
|
|
8
|
+
- "tamper test"
|
|
9
|
+
- "placement discipline"
|
|
4
10
|
validated: 2026-09-15
|
|
5
11
|
verdict: UNVALIDATED
|
|
6
12
|
---
|
|
7
13
|
|
|
8
|
-
## trigger: "placement, wrong layer, generic layer, domain fact in shared code, ## Placement section, placement ledger, tamper test, validating a worker result"
|
|
9
|
-
|
|
10
14
|
# Placement Discipline
|
|
11
15
|
|
|
12
16
|
A fix lives where the invariant lives: shared layers know mechanisms, never one project's names, paths, or policy. When a change would teach a generic layer a domain fact, the design is wrong - stop and move the mechanism, do not patch around it.
|
|
@@ -1,12 +1,16 @@
|
|
|
1
1
|
---
|
|
2
2
|
name: probe-evaluation
|
|
3
3
|
description: "Design and score a behavioral A/B probe so a null result is informative: pre-registered rubric and decision rule, control-arm sanity, blinded scoring, power and statistics (Wilson intervals, Fisher exact), and the failure catalog for probes that measured nothing. Use when probing whether a prompt, extension, or config change causes a behavioral delta, when extending a probe matrix, or when judging whether a probe's null vetoes or clears a change."
|
|
4
|
+
triggers:
|
|
5
|
+
- "behavioral probe"
|
|
6
|
+
- "a/b probe"
|
|
7
|
+
- "pre-registered rubric"
|
|
8
|
+
- "probe came back null"
|
|
9
|
+
- "two-arm comparison"
|
|
4
10
|
validated: 2026-09-15
|
|
5
11
|
verdict: UNVALIDATED
|
|
6
12
|
---
|
|
7
13
|
|
|
8
|
-
## trigger: "behavioral probe, A/B probe, prompt A/B test, probe came back null, scoring probe runs, pre-registered rubric, two-arm comparison"
|
|
9
|
-
|
|
10
14
|
# Probe Evaluation
|
|
11
15
|
|
|
12
16
|
A probe answers one question: does arm B (treatment) behave differently from arm A (control)? The ship-gate rule: **no directional delta under the pre-registered decision rule means the change does not ship.** The symmetric rule: a null whose own preconditions failed (control never showed the failure, judge unblinded, N too small) demonstrates nothing in either direction. "No demonstrated benefit" and "demonstrated harm" are different verdicts; never flatten them.
|
|
@@ -1,7 +1,11 @@
|
|
|
1
1
|
---
|
|
2
2
|
name: release-pipeline
|
|
3
|
-
description: "
|
|
4
|
-
|
|
3
|
+
description: "Use when releasing braid packages: local builds, packing, installs, and verification; remote npm publication through GitHub Actions OIDC."
|
|
4
|
+
triggers:
|
|
5
|
+
- "release pipeline"
|
|
6
|
+
- "npm publish"
|
|
7
|
+
- "npm pack"
|
|
8
|
+
- "release worktree"
|
|
5
9
|
validated: 2026-10-18
|
|
6
10
|
verdict: IMPROVE
|
|
7
11
|
---
|
package/skills/research/SKILL.md
CHANGED
|
@@ -1,6 +1,12 @@
|
|
|
1
1
|
---
|
|
2
2
|
name: research
|
|
3
3
|
description: "Research topics, verify facts, and find information online. Covers web search (DuckDuckGo HTML as primary, SearXNG as unreliable fallback), direct page fetching, fact-checking, literature review, and evidence gathering. Use when looking up information, verifying claims, comparing sources, or finding current data."
|
|
4
|
+
triggers:
|
|
5
|
+
- "research"
|
|
6
|
+
- "search the web"
|
|
7
|
+
- "look this up"
|
|
8
|
+
- "verify a fact"
|
|
9
|
+
- "literature review"
|
|
4
10
|
validated: 2026-08-16
|
|
5
11
|
verdict: KEEP
|
|
6
12
|
---
|
|
@@ -1,6 +1,12 @@
|
|
|
1
1
|
---
|
|
2
2
|
name: skill-usage
|
|
3
3
|
description: "Measure skill adoption from durable counters and session telemetry: per-skill suggested/loaded/claimed/actual counts, conversion rates, and active/needs-refresh/prune buckets. Use when measuring whether skills are actually used, taking before/after metrics around a skill or scoring change, deciding keep/refresh/deprecate for a skill, or reporting adoption - the skills-subsystem instrument (skill_stats tool, skills-usage.json, injection-tracker). Do NOT use for general research, code search, or project metrics outside the skills subsystem."
|
|
4
|
+
triggers:
|
|
5
|
+
- "skill adoption"
|
|
6
|
+
- "skill usage"
|
|
7
|
+
- "skill_stats"
|
|
8
|
+
- "needs-refresh"
|
|
9
|
+
- "adoption report"
|
|
4
10
|
validated: 2026-09-09
|
|
5
11
|
verdict: KEEP
|
|
6
12
|
---
|
|
@@ -1,12 +1,15 @@
|
|
|
1
1
|
---
|
|
2
2
|
name: software-engineer
|
|
3
3
|
description: "Use when writing, reviewing, or improving code. Applies universally: debugging (understand → trace → fix → verify), implementation (analyze → decompose → execute), and code review (read → critique → suggest). The universal protocol works for any language, framework, or task size."
|
|
4
|
+
triggers:
|
|
5
|
+
- "writing code"
|
|
6
|
+
- "reviewing code"
|
|
7
|
+
- "debugging"
|
|
8
|
+
- "code review"
|
|
4
9
|
validated: 2026-08-16
|
|
5
10
|
verdict: KEEP
|
|
6
11
|
---
|
|
7
12
|
|
|
8
|
-
## trigger: "writing code, reviewing code, improving code, debugging, implementation, code review, refactoring, any language, any framework"
|
|
9
|
-
|
|
10
13
|
# Software Engineer
|
|
11
14
|
|
|
12
15
|
Universal protocol for any coding task: analyze, decompose, execute, verify. Debugging follows the same path but starts with understanding the failure.
|
|
@@ -1,7 +1,12 @@
|
|
|
1
1
|
---
|
|
2
2
|
name: subagent-orchestration
|
|
3
|
-
description: "
|
|
4
|
-
|
|
3
|
+
description: "Use when coordinating multi-agent work: spawn workers, manage lifecycles, signal between agents, transfer context, handle errors, and cascade to fresh sessions. Covers both orchestrator and worker roles - an orchestrator is a subagent that owns coordination; every subagent orchestrates its own tools."
|
|
4
|
+
triggers:
|
|
5
|
+
- "spawn a worker"
|
|
6
|
+
- "spawn workers"
|
|
7
|
+
- "multi-agent"
|
|
8
|
+
- "mastership transfer"
|
|
9
|
+
- "context handoff"
|
|
5
10
|
validated: 2026-08-16
|
|
6
11
|
verdict: KEEP
|
|
7
12
|
---
|
|
@@ -1,7 +1,12 @@
|
|
|
1
1
|
---
|
|
2
2
|
name: subagent-verification
|
|
3
3
|
description: "Verify changes by spawning an isolated subagent before merging. Use when: modifying braid core or contrib, running tests after edits, validating that an integration doesn't break existing functionality. Covers: spawn-and-verify workflow, test-gate integration, failure diagnosis, and merge gates."
|
|
4
|
-
|
|
4
|
+
triggers:
|
|
5
|
+
- "verify worker"
|
|
6
|
+
- "isolated subagent"
|
|
7
|
+
- "verifier subagent"
|
|
8
|
+
- "verify before merging"
|
|
9
|
+
- "spawn a verifier"
|
|
5
10
|
validated: 2026-08-16
|
|
6
11
|
verdict: IMPROVE
|
|
7
12
|
---
|
|
@@ -1,12 +1,15 @@
|
|
|
1
1
|
---
|
|
2
2
|
name: tone-quality
|
|
3
|
-
description: "
|
|
3
|
+
description: "Use when writing or editing text that humans or agents will read - documentation, prompts, messages, skills, comments, and explanations. Read this skill before writing and ask yourself: 'Am I writing meaningful text?' If yes, activate this skill. Apply these principles while drafting; this is not a post-hoc review step. Covers directive density, cognitive reasoning density, voice artifacts, epistemic calibration, and signal density."
|
|
4
|
+
triggers:
|
|
5
|
+
- "write the docs"
|
|
6
|
+
- "write documentation"
|
|
7
|
+
- "draft text"
|
|
8
|
+
- "edit text"
|
|
4
9
|
validated: 2026-08-16
|
|
5
10
|
verdict: KEEP
|
|
6
11
|
---
|
|
7
12
|
|
|
8
|
-
## trigger: "writing documentation, writing prompts, writing messages, writing skills, writing comments, writing explanations, drafting text, editing text"
|
|
9
|
-
|
|
10
13
|
# Tone Quality
|
|
11
14
|
|
|
12
15
|
Apply four diagnostic dimensions when creating or revising text. Each dimension catches problems the others miss.
|
|
@@ -1,7 +1,12 @@
|
|
|
1
1
|
---
|
|
2
2
|
name: voice-replies
|
|
3
|
-
description: "Reply to Telegram voice messages with synthesized speech: edge-tts (RU/EN/HE voices) → ffmpeg ogg/opus → Bot API sendVoice, matching the incoming message's language. Use when a voice message arrived and the reply should be a voice note in the same language, or when asked to send a Telegram voice reply.
|
|
4
|
-
|
|
3
|
+
description: "Reply to Telegram voice messages with synthesized speech: edge-tts (RU/EN/HE voices) → ffmpeg ogg/opus → Bot API sendVoice, matching the incoming message's language. Use when a voice message arrived and the reply should be a voice note in the same language, or when asked to send a Telegram voice reply. Includes offline fallbacks (piper, macOS say)."
|
|
4
|
+
triggers:
|
|
5
|
+
- "voice reply"
|
|
6
|
+
- "voice message"
|
|
7
|
+
- "sendVoice"
|
|
8
|
+
- "text-to-speech"
|
|
9
|
+
- "edge-tts"
|
|
5
10
|
validated: 2026-11-03
|
|
6
11
|
verdict: UNVALIDATED
|
|
7
12
|
---
|