diff --git a/.claude/skills/piker-profiling/SKILL.md b/.agents/skills/piker-profiling/SKILL.md
similarity index 100%
rename from .claude/skills/piker-profiling/SKILL.md
rename to .agents/skills/piker-profiling/SKILL.md
diff --git a/.claude/skills/piker-profiling/patterns.md b/.agents/skills/piker-profiling/patterns.md
similarity index 100%
rename from .claude/skills/piker-profiling/patterns.md
rename to .agents/skills/piker-profiling/patterns.md
diff --git a/.claude/skills/piker-slang/SKILL.md b/.agents/skills/piker-slang/SKILL.md
similarity index 100%
rename from .claude/skills/piker-slang/SKILL.md
rename to .agents/skills/piker-slang/SKILL.md
diff --git a/.claude/skills/piker-slang/dictionary.md b/.agents/skills/piker-slang/dictionary.md
similarity index 100%
rename from .claude/skills/piker-slang/dictionary.md
rename to .agents/skills/piker-slang/dictionary.md
diff --git a/.claude/skills/piker-slang/examples.md b/.agents/skills/piker-slang/examples.md
similarity index 100%
rename from .claude/skills/piker-slang/examples.md
rename to .agents/skills/piker-slang/examples.md
diff --git a/.claude/skills/pyqtgraph-optimization/SKILL.md b/.agents/skills/pyqtgraph-optimization/SKILL.md
similarity index 100%
rename from .claude/skills/pyqtgraph-optimization/SKILL.md
rename to .agents/skills/pyqtgraph-optimization/SKILL.md
diff --git a/.claude/skills/pyqtgraph-optimization/examples.md b/.agents/skills/pyqtgraph-optimization/examples.md
similarity index 100%
rename from .claude/skills/pyqtgraph-optimization/examples.md
rename to .agents/skills/pyqtgraph-optimization/examples.md
diff --git a/.claude/skills/timeseries-optimization/SKILL.md b/.agents/skills/timeseries-optimization/SKILL.md
similarity index 100%
rename from .claude/skills/timeseries-optimization/SKILL.md
rename to .agents/skills/timeseries-optimization/SKILL.md
diff --git a/.claude/skills/timeseries-optimization/numpy-patterns.md b/.agents/skills/timeseries-optimization/numpy-patterns.md
similarity index 100%
rename from .claude/skills/timeseries-optimization/numpy-patterns.md
rename to .agents/skills/timeseries-optimization/numpy-patterns.md
diff --git a/.claude/skills/timeseries-optimization/polars-patterns.md b/.agents/skills/timeseries-optimization/polars-patterns.md
similarity index 100%
rename from .claude/skills/timeseries-optimization/polars-patterns.md
rename to .agents/skills/timeseries-optimization/polars-patterns.md
diff --git a/.claude/settings.local.json b/.claude/settings.local.json
deleted file mode 100644
index 5404a5dd..00000000
--- a/.claude/settings.local.json
+++ /dev/null
@@ -1,11 +0,0 @@
-{
-  "permissions": {
-    "allow": [
-      "Bash(chmod:*)",
-      "Bash(/tmp/piker_commits.txt)",
-      "Bash(python:*)"
-    ],
-    "deny": [],
-    "ask": []
-  }
-}
diff --git a/.claude/skills/commit-msg/style-guide-reference.md b/.claude/skills/commit-msg/style-guide-reference.md
deleted file mode 100644
index 1e3ba061..00000000
--- a/.claude/skills/commit-msg/style-guide-reference.md
+++ /dev/null
@@ -1,262 +0,0 @@
-# Piker Git Commit Message Style Guide
-
-Learned from analyzing 500 commits from the piker repository.
-
-## Subject Line Rules
-
-### Length
-- Target: ~50 characters (avg: 50.5 chars)
-- Maximum: 67 chars (hard limit, though historical max: 146)
-- Keep concise and descriptive
-
-### Structure
-- Use present tense verbs (Add, Drop, Fix, Move, etc.)
-- 65.6% of commits use backticks for code references
-- 33.0% use colon notation (`module.file:` prefix or `: ` separator)
-
-### Opening Verbs (by frequency)
-Primary verbs to use:
-- **Add** (8.4%) - New features, files, functionality
-- **Drop** (3.2%) - Remove features, dependencies, code
-- **Fix** (2.2%) - Bug fixes, corrections
-- **Use** (2.2%) - Switch to different approach/tool
-- **Port** (2.0%) - Migrate code, adapt from elsewhere
-- **Move** (2.0%) - Relocate code, refactor structure
-- **Always** (1.8%) - Enforce consistent behavior
-- **Factor** (1.6%) - Refactoring, code organization
-- **Bump** (1.6%) - Version/dependency updates
-- **Update** (1.4%) - Modify existing functionality
-- **Adjust** (1.0%) - Fine-tune, tweak behavior
-- **Change** (1.0%) - Modify behavior or structure
-
-Casual/informal verbs (used occasionally):
-- **Woops,** (1.4%) - Fixing mistakes
-- **Lul,** (0.6%) - Humorous corrections
-
-### Code References
-Use backticks heavily for:
-- **Module/package names**: `tractor`, `pikerd`, `polars`, `ruff`
-- **Data types**: `dict`, `float`, `str`, `None`
-- **Classes**: `MktPair`, `Asset`, `Position`, `Account`, `Flume`
-- **Functions**: `dedupe()`, `push()`, `get_client()`, `norm_trade()`
-- **File paths**: `.tsp`, `.fqme`, `brokers.toml`, `conf.toml`
-- **CLI flags**: `--pdb`
-- **Error types**: `NoData`
-- **Tools**: `uv`, `uv sync`, `httpx`, `numpy`
-
-### Colon Usage Patterns
-1. **Module prefix**: `.ib.feed: trim bars frame to start_dt`
-2. **Separator**: `Add support: new feature description`
-
-### Tone
-- Technical but casual (use XD, lol, .., Woops, Lul when appropriate)
-- Direct and concise
-- Question marks rare (1.4%)
-- Exclamation marks rare (1.4%)
-
-## Body Structure
-
-### Body Frequency
-- 56.0% of commits have empty bodies (one-line commits are common)
-- Use body for complex changes requiring explanation
-
-### Bullet Lists
-- Prefer `-` bullets (16.2% of commits)
-- Rarely use `*` bullets (1.6%)
-- Indent continuation lines appropriately
-
-### Section Markers (in order of frequency)
-Use these to organize complex commit bodies:
-
-1. **Also,** (most common, 26 occurrences)
-   - Additional changes, side effects, related updates
-   - Example:
-     ```
-     Main change described in subject.
-
-     Also,
-     - related change 1
-     - related change 2
-     ```
-
-2. **Deats,** (8 occurrences)
-   - Implementation details
-   - Technical specifics
-
-3. **Further,** (4 occurrences)
-   - Additional context or future considerations
-
-4. **Other,** (3 occurrences)
-   - Miscellaneous related changes
-
-5. **Notes,** **TODO,** (rare, 1 each)
-   - Special annotations when needed
-
-### Line Length
-- Body lines: 67 character maximum
-- Break longer lines appropriately
-
-## Language Patterns
-
-### Common Abbreviations (by frequency)
-Use these freely in commit bodies:
-- **msg** (29) - message
-- **mod** (15) - module
-- **vs** (14) - versus
-- **impl** (12) - implementation
-- **deps** (11) - dependencies
-- **var** (6) - variable
-- **ctx** (6) - context
-- **bc** (5) - because
-- **obvi** (4) - obviously
-- **ep** (4) - endpoint
-- **tn** (4) - task name
-- **rn** (3) - right now
-- **sig** (3) - signal/signature
-- **env** (3) - environment
-- **tho** (3) - though
-- **fn** (2) - function
-- **iface** (2) - interface
-- **prolly** (2) - probably
-
-Less common but acceptable:
-- **dne**, **osenv**, **gonna**, **wtf**
-
-### Tone Indicators
-- **..** (77 occurrences) - Ellipsis for trailing thoughts
-- **XD** (17) - Expression of humor/irony
-- **lol** (1) - Rare, use sparingly
-
-### Informal Patterns
-- Casual contractions okay: Don't, won't
-- Lowercase starts acceptable for file prefixes
-- Direct, conversational tone
-
-## Special Patterns
-
-### Module/File Prefixes
-Common in piker commits (33.0% use colons):
-- `.ib.feed: description`
-- `.ui._remote_ctl: description`
-- `.data.tsp: description`
-- `.accounting: description`
-
-### Merge Commits
-- 4.4% of commits (standard git merges)
-- Not a primary pattern to emulate
-
-### External References
-- GitHub links occasionally used (13 total)
-- File:line references not used (0 occurrences)
-- No WIP commits in analyzed set
-
-### Claude-code Footer
-When the written **patch** was assisted by claude-code,
-include:
-
-```
-(this patch was generated in some part by [`claude-code`][claude-code-gh])
-[claude-code-gh]: https://github.com/anthropics/claude-code
-```
-
-When only the **commit msg** was written by claude-code
-(human wrote the patch), use:
-
-```
-(this commit msg was generated in some part by [`claude-code`][claude-code-gh])
-[claude-code-gh]: https://github.com/anthropics/claude-code
-```
-
-## Piker-Specific Terms
-
-### Core Components
-- `pikerd` - piker daemon
-- `brokerd` - broker daemon
-- `tractor` - actor framework used
-- `.tsp` - time series protocol/module
-- `.fqme` - fully qualified market endpoint
-
-### Data Structures
-- `MktPair` - market pair
-- `Asset` - asset representation
-- `Position` - trading position
-- `Account` - account data
-- `Flume` - data stream
-- `SymbologyCache` - symbol caching
-
-### Common Functions
-- `dedupe()` - deduplication
-- `push()` - data pushing
-- `get_client()` - client retrieval
-- `norm_trade()` - trade normalization
-- `open_trade_ledger()` - ledger opening
-- `markup_gaps()` - gap marking
-- `get_null_segs()` - null segment retrieval
-- `remote_annotate()` - remote annotation
-
-### Brokers & Integrations
-- `binance` - Binance integration
-- `.ib` - Interactive Brokers
-- `bs_mktid` - broker-specific market ID
-- `reqid` - request ID
-
-### Configuration
-- `brokers.toml` - broker configuration
-- `conf.toml` - general configuration
-
-### Development Tools
-- `ruff` - Python linter
-- `uv` / `uv sync` - package manager
-- `--pdb` - debugger flag
-- `pdbp` - debugger
-- `asyncvnc` / `pyvnc` - VNC libraries
-- `httpx` - HTTP client
-- `polars` - dataframe library
-- `rapidfuzz` - fuzzy matching
-- `numpy` - numerical library
-- `trio` - async framework
-- `asyncio` - async framework
-- `xonsh` - shell
-
-## Examples
-
-### Simple one-liner
-```
-Add `MktPair.fqme` property for symbol resolution
-```
-
-### With module prefix
-```
-.ib.feed: trim bars frame to `start_dt`
-```
-
-### Casual fix
-```
-Woops, compare against first-dt in `.ib.feed` bars frame
-```
-
-### With body using "Also,"
-```
-Drop `poetry` for `uv` in dev workflow
-
-Also,
-- update deps in `pyproject.toml`
-- add `uv sync` to CI pipeline
-- remove old `poetry.lock`
-```
-
-### With implementation details
-```
-Factor position tracking into `Position` dataclass
-
-Deats,
-- move calc logic from `brokerd` to `.accounting`
-- add `norm_trade()` helper for broker normalization
-- use `MktPair.fqme` for consistent symbol refs
-```
-
----
-
-**Analysis date:** 2026-01-27
-**Commits analyzed:** 500 from piker repository
-**Maintained by:** Tyler Goodlet
diff --git a/.claude/skills/piker-profiling b/.claude/skills/piker-profiling
new file mode 120000
index 00000000..2b3636a0
--- /dev/null
+++ b/.claude/skills/piker-profiling
@@ -0,0 +1 @@
+../../.agents/skills/piker-profiling
\ No newline at end of file
diff --git a/.claude/skills/piker-slang b/.claude/skills/piker-slang
new file mode 120000
index 00000000..85f9cbec
--- /dev/null
+++ b/.claude/skills/piker-slang
@@ -0,0 +1 @@
+../../.agents/skills/piker-slang
\ No newline at end of file
diff --git a/.claude/skills/pyqtgraph-optimization b/.claude/skills/pyqtgraph-optimization
new file mode 120000
index 00000000..b1ca4594
--- /dev/null
+++ b/.claude/skills/pyqtgraph-optimization
@@ -0,0 +1 @@
+../../.agents/skills/pyqtgraph-optimization
\ No newline at end of file
diff --git a/.claude/skills/timeseries-optimization b/.claude/skills/timeseries-optimization
new file mode 120000
index 00000000..2bd3abd1
--- /dev/null
+++ b/.claude/skills/timeseries-optimization
@@ -0,0 +1 @@
+../../.agents/skills/timeseries-optimization
\ No newline at end of file
diff --git a/ai/prompt-io/opencode/20260908T043134Z_f148fbcc_prompt_io.md b/ai/prompt-io/opencode/20260908T043134Z_f148fbcc_prompt_io.md
new file mode 100644
index 00000000..95251360
--- /dev/null
+++ b/ai/prompt-io/opencode/20260908T043134Z_f148fbcc_prompt_io.md
@@ -0,0 +1,37 @@
+---
+model: openai/gpt-5.6-sol
+service: opencode
+session: unavailable
+timestamp: 2026-09-08T04:31:34Z
+git_ref: f148fbcc
+scope: config
+substantive: true
+raw_file: 20260908T043134Z_f148fbcc_prompt_io.raw.md
+---
+
+## Prompt
+
+Audit tracked legacy skill files under `.claude/`, then migrate Piker's
+repository-specific skills into the provider-neutral discovery tree.
+Keep Claude compatibility and remove stale Claude-local configuration.
+
+## Response summary
+
+Moves four Piker-specific skill packages to `.agents/skills/`, retains
+Claude discovery through relative adapter links, and drops two stale
+tracked Claude-local files.
+
+## Files changed
+
+- `.agents/skills/piker-*/` - provider-neutral Piker skill sources.
+- `.claude/skills/piker-*` - relative Claude compatibility links.
+- `.claude/settings.local.json` - remove tracked local permissions.
+- `.claude/skills/commit-msg/style-guide-reference.md` - remove the
+  superseded legacy copy now represented under `.ai/commit-msg/`.
+
+## Human edits
+
+The human identified that repository-specific skills were not migrated
+by the shared `ai.skillz` deployment, selected `.agents/skills/` as the
+neutral source, and requested that the migration land in PR 93 before
+following up on the generic migration tooling.
diff --git a/ai/prompt-io/opencode/20260908T043134Z_f148fbcc_prompt_io.raw.md b/ai/prompt-io/opencode/20260908T043134Z_f148fbcc_prompt_io.raw.md
new file mode 100644
index 00000000..3af350df
--- /dev/null
+++ b/ai/prompt-io/opencode/20260908T043134Z_f148fbcc_prompt_io.raw.md
@@ -0,0 +1,34 @@
+---
+model: openai/gpt-5.6-sol
+service: opencode
+timestamp: 2026-09-08T04:31:34Z
+git_ref: f148fbcc
+diff_cmd: git diff --cached
+---
+
+## Prompt
+
+Audit tracked legacy skill files under `.claude/`, then migrate Piker's
+repository-specific skills into the provider-neutral discovery tree.
+Keep Claude compatibility and remove stale Claude-local configuration.
+
+## Generated changes
+
+> `git diff --cached -- .agents/skills/`
+
+Moves the four Piker-owned skill packages into the neutral
+`.agents/skills/` source tree without changing their contents.
+
+> `git diff --cached -- .claude/skills/ .claude/settings.local.json`
+
+Adds relative Claude adapter links to the neutral skill sources and
+removes the tracked local settings file and superseded commit-message
+style guide.
+
+## Validation
+
+- All four Claude adapter links resolve to a `SKILL.md`.
+- `git diff --cached --check` passes.
+- No tracked files match the repository's ignore rules.
+- Shared deployment validation retains one unrelated stale local
+  `commit-plan` hybrid-link error for an `ai.skillz` follow-up.