@event4u/agent-config 3.3.0 → 4.2.0

package/docs/contracts/iron-law-overrides.txt

@@ -15,7 +15,7 @@
 # ADR-002 (2026-05-06) — kernel-bucket cap raise + per-rule overrides.
 # Source: docs/decisions/ADR-002-kernel-bucket-overrides.md
 # Empirical r_actual=0.795 vs r_projected=0.712. The 6 rules below carry
-# Iron-Law fences + non-externalisable enumerations; further compression
+# Iron-Law fences + non-externalisable enumerations; further condensation
 # would force semantic loss. All sit ≤ 4 000-char override ceiling.
 ask-when-uncertain
 direct-answers

package/docs/contracts/kernel-membership.md

@@ -10,7 +10,7 @@ keep-beta-until: 2026-08-12
 > · **Source of truth:** `docs/contracts/rule-classification.md` § 3.1
 > · **Pilot artifacts:** `docs/contracts/pilot/`
 
-Locks the kernel set and the empirical compression rate `r` derived
+Locks the kernel set and the empirical condensation rate `r` derived
 from the P1.3 pilot. Phase 2.2 executes against this contract.
 
 ## § 1 — Inclusion criteria (kernel-only)
@@ -37,7 +37,7 @@ A rule belongs in the kernel iff it satisfies **all** of:
    `no-cheap-questions` are first-class kernel members even though
    they fire during deliberation, not strictly pre-send.)*
 
-All other rules → `compress-and-keep` (auto-tier) or migrate to
+All other rules → `condense-and-keep` (auto-tier) or migrate to
 skill / guideline per `rule-classification.md`.
 
 > **Council amendment (P1.4).** Criterion #3 originally read "must be
@@ -49,12 +49,12 @@ skill / guideline per `rule-classification.md`.
 > existing 9-rule kernel set. See
 > `agents/runtime/council/sessions/20260506T044941Z-phase1-cross-check-r2.json`. <!-- council-ref-allowed: contract decision trace -->
 
-## § 2 — Empirical compression rate (pilot)
+## § 2 — Empirical condensation rate (pilot)
 
 Pilot picks (per P1.3 algorithm: shortest / median / longest of the 9
 always-rules in `rule-classification.md` § 3.1):
 
-| pilot | original | compressed | r | per-rule cap (2.5k) | Iron-Law SHA |
+| pilot | original | condensed | r | per-rule cap (2.5k) | Iron-Law SHA |
 |---|---:|---:|---:|---|---|
 | agent-authority   | 1217 | 1020 | 0.838 | OK        | match (`e3b0c442…`) |
 | direct-answers    | 3991 | 2841 | 0.712 | +341 over | match (`ef438530…`) |
@@ -65,21 +65,21 @@ always-rules in `rule-classification.md` § 3.1):
 | mean r (outlier-skewed by lean pilot) | 0.742 |
 | **median r** (locked for projection) | **0.712** |
 | max r (already-lean floor) | 0.838 |
-| min r (most compressible) | 0.677 |
+| min r (most condensable) | 0.677 |
 
 Locked `r = 0.712` (median) for P2.2 projection. Council amendment
 (P1.4) — the original lock was the **mean** (0.742), but Sonnet 4.5
 flagged outlier skew: `agent-authority` at 1217 chars pre-pilot was
-already near the compression floor (`r = 0.838`), pulling the mean
+already near the condensation floor (`r = 0.838`), pulling the mean
 above the typical band. Median (0.712) is robust to the lean
 outlier and sits inside the roadmap-stated 0.6–0.75 band. Risk
-asymmetry confirms the choice: under-estimating compressed size
+asymmetry confirms the choice: under-estimating condensed size
 blows the 25k cap mid-P2.2 (no abort path); over-estimating just
 yields headroom. GPT-4o concurred. Source:
 `agents/runtime/council/sessions/20260506T044941Z-phase1-cross-check-r2.json`. <!-- council-ref-allowed: contract decision trace -->
 
 **Pilot caveat — already-lean floor.** `agent-authority` (1217 chars
-pre-pilot) yielded only `r = 0.838`. Pre-compressed rules give less
+pre-pilot) yielded only `r = 0.838`. Pre-condensed rules give less
 back; rules > 3k chars give more back. Future kernel additions
 ≤ 1.5k chars should not assume the median — they may already be at
 floor.
@@ -117,7 +117,7 @@ absorbs pilot variance for the 6 untested rules.
 
 ### § 4.1 — Post-P2.2 locked kernel (empirical, 2026-05-06)
 
-P2.2 applied compression (imperative rewrite, examples → guidelines,
+P2.2 applied condensation (imperative rewrite, examples → guidelines,
 rationale → contexts) + verified Iron-Law fence preservation.
 Empirical r=0.795 (vs 0.712 projected). Bucket cap raised 25k → 26k
 via `docs/decisions/ADR-002-kernel-bucket-overrides.md`. Six rules
@@ -164,7 +164,7 @@ until the next kernel re-measurement.
 
 † **agent-authority swap candidate (P1.4 ADR).** Sonnet 4.5 argues
 this is a routing index (zero Iron Law fences, dispatches to other
-kernel rules) and should be `compress-and-keep` (auto-tier-3),
+kernel rules) and should be `condense-and-keep` (auto-tier-3),
 freeing 867 projected chars. The Council also recommends promoting
 `autonomous-execution` (5631 → 4009 projected) into kernel as a
 mode-independent Band-4 authority. Net effect: 23071 − 867 + 4009 =
@@ -191,10 +191,10 @@ Both Council members concurred — the original 1.5k figure forced 8 of
 The 2.5k cap fits 7 of 9 rules without ADR; only the two largest
 remain as legitimate Iron-Law-density exceptions.
 
-Rules that land **above the 2.5k per-rule cap** even after compression
+Rules that land **above the 2.5k per-rule cap** even after condensation
 and Iron-Law preservation:
 
-| rule | post-compression chars | over-cap (2.5k) | reason cap cannot be hit |
+| rule | post-condensation chars | over-cap (2.5k) | reason cap cannot be hit |
 |---|---:|---:|---|
 | direct-answers    | 2842 | +342  | Three Iron-Law fences (≈ 470 chars) + severity-tiered claim table required for invented-facts gate |
 | language-and-tone | 3786 | +1286 | Iron-Law fence + exhaustive user-visible-prose catalog + pre-send gate are co-load-bearing |
@@ -204,12 +204,12 @@ and Iron-Law preservation:
 ADR (see § 5.2).
 
 Per roadmap § "Iron-Law over budget = ADR exception, not auto-
-compress" — these surface as `iron-law-override` exceptions for P2.1
+condense" — these surface as `iron-law-override` exceptions for P2.1
 to gate behind ADRs in `docs/decisions/`.
 
 The remaining 6 always-rules project to 1669–2916 chars post-
-compression — all under the 2.5k cap with no ADR needed. P2.2
-measures the real post-compression sizes; if any breach the cap, an
+condensation — all under the 2.5k cap with no ADR needed. P2.2
+measures the real post-condensation sizes; if any breach the cap, an
 ADR lands per-rule.
 
 ### § 5.2 — Kernel-set swap ADR (agent-authority ↔ autonomous-execution)
@@ -236,19 +236,19 @@ clearing the way for the swap under a fresh ADR.
 
 ## § 6 — Abort criteria for P2.2
 
-**Council amendment (P1.4).** Sonnet 4.5 flagged that P2.2 compresses
+**Council amendment (P1.4).** Sonnet 4.5 flagged that P2.2 condenses
 in place with no defined abort path. The following triggers force
 revert + re-projection rather than mid-phase scramble:
 
 - **Iron-Law drift.** Any kernel rule loses Iron-Law SHA equivalence
-  after compression (whitespace-normalised, case-folded fence-block
-  content). Single occurrence aborts the rule's compression.
-- **Bucket overflow.** Kernel sum post-compression > 27 500 chars
+  after condensation (whitespace-normalised, case-folded fence-block
+  content). Single occurrence aborts the rule's condensation.
+- **Bucket overflow.** Kernel sum post-condensation > 27 500 chars
   (10 % over hard cap). Aborts the entire P2.2 phase.
-- **Single-rule runaway.** Any kernel rule post-compression
-  > 4 000 chars. Aborts that rule's compression; flag for ADR.
+- **Single-rule runaway.** Any kernel rule post-condensation
+  > 4 000 chars. Aborts that rule's condensation; flag for ADR.
 - **Empirical r drift.** Median of measured r across the first 4
-  compressed rules deviates > 0.10 from the locked 0.712 (i.e.,
+  condensed rules deviates > 0.10 from the locked 0.712 (i.e.,
   measured median outside [0.612, 0.812]). P2.2 pauses, re-locks r,
   re-runs § 3 projection.
 
@@ -266,10 +266,10 @@ with the trigger-relaxed parameter, lock new value, re-attempt.
 
 ## § 8 — What this contract does NOT do
 
-- It does **not** edit any source rule. P2.2 ships compression to
-  `.agent-src.uncompressed/rules/`.
+- It does **not** edit any source rule. P2.2 ships condensation to
+  `.agent-src.uncondensed/rules/`.
 - It does **not** lock per-rule `chars` (only the bucket sum). Per-
-  rule values are projections; P2.2 measures real post-compression.
+  rule values are projections; P2.2 measures real post-condensation.
 - It does **not** decide kernel additions. Adding a rule to the
   kernel requires an ADR per roadmap § "Sequencing-drift time-box"
   and re-runs P1.3 with the new candidate.
@@ -288,4 +288,4 @@ with the trigger-relaxed parameter, lock new value, re-attempt.
 | 2026-05-06 | `agents/runtime/council/sessions/20260506T044821Z-phase1-cross-check.json` | Council R1 (truncated at 1024 tokens) | <!-- council-ref-allowed: contract decision trace -->
 | 2026-05-06 | `agents/runtime/council/sessions/20260506T044941Z-phase1-cross-check-r2.json` | Council R2 (3500 tokens) — locks median r, splits criterion #3, adds criterion #5, raises per-rule cap to 2.5k, defines abort criteria | <!-- council-ref-allowed: contract decision trace -->
 | 2026-05-06 | this file | kernel set locked: 9 rules, projected 23 071 chars (median r = 0.712) |
-| 2026-05-06 | P2.2 compression + `scripts/iron_law_sha.py --all-kernel` | empirical: 25 590 chars (r_actual = 0.795 across kernel; longer rules compress less than the pilot median) — see `docs/decisions/ADR-002-kernel-bucket-overrides.md` (KERNEL_HARD raised 25k → 26k, 6 per-rule overrides ≤ 4k ceiling, all Iron-Law SHAs preserved) |
+| 2026-05-06 | P2.2 condensation + `scripts/iron_law_sha.py --all-kernel` | empirical: 25 590 chars (r_actual = 0.795 across kernel; longer rules condense less than the pilot median) — see `docs/decisions/ADR-002-kernel-bucket-overrides.md` (KERNEL_HARD raised 25k → 26k, 6 per-rule overrides ≤ 4k ceiling, all Iron-Law SHAs preserved) |

package/docs/contracts/linear-ai-rules-inclusion.md

@@ -130,7 +130,7 @@ persistence on cloud platforms; the rule is a no-op.
   partitions workspace / team / personal. Step 3 only needs to
   document the three-layer rationale for the README.
 - Re-classify on every audit run; if a rule is added to
-  `.agent-src.uncompressed/rules/`, this document MUST gain a row.
+  `.agent-src.uncondensed/rules/`, this document MUST gain a row.
 
 ## Source of tier data
 

package/docs/contracts/linter-structural-model.md

@@ -88,7 +88,7 @@ procedure → no warning under the new model. ✓ roadmap success criterion.
 success criterion.
 
 Of 13 commands ≥ 1000 words, only **2** lack a delegation signal —
-both are candidates for Phase 4.1 review (`compress.md`,
+both are candidates for Phase 4.1 review (`condense.md`,
 `project-analyze.md`; the latter has density 0.86, exempt under the
 density-AND-delegation gate).
 
@@ -115,7 +115,7 @@ Calibration sweep on the 2026-05-08 corpus (310 artifacts):
 Pass rate: 186 → 209 (`pass`); 124 → 101 (`pass_with_warnings`); 0
 errors. Each remaining warning is a genuine structural defect:
 
-- `compress.md` (1569 words, density 0.58, no delegation signal) —
+- `condense.md` (1569 words, density 0.58, no delegation signal) —
   inlined logic in a non-orchestrator command.
 - `artifact-drafting-protocol.md` rule (65 lines, density 0.37, no
   Iron-Law block) — prose-dominant long rule.

package/docs/contracts/load-context-budget-model.md

@@ -44,10 +44,10 @@ Where:
   (per [`load-context-schema.md`](load-context-schema.md)).
 - `transitive_closure` walks `load_context:` declarations on
   context files **up to depth 2** — see § Nesting cap below.
-- `RawSize` is the byte size of the compressed file
+- `RawSize` is the byte size of the condensed file
   (`.agent-src/...`), measured by `os.path.getsize()`. The
-  uncompressed source paths in frontmatter are mapped to their
-  compressed counterparts before sizing.
+  uncondensed source paths in frontmatter are mapped to their
+  condensed counterparts before sizing.
 - A context loaded by N always-rules counts **N times** (once per
   loading rule). Rationale: the always-budget protects context-window
   utilization at activation time; if rule A and rule B both fire on
@@ -138,7 +138,7 @@ the full `RawSize(X)` under Model (b). No `chars(X) / N` discount.
   has no measurable upside.
 - Phase 4c (shared-context discount) becomes a no-op under 3a.
   Phase 4 leverage shifts to 4a (demote), 4b (merge), and 4d
-  (hard-compress) — see `road-to-context-layer-maturity` Phase 4
+  (hard-condense) — see `road-to-context-layer-maturity` Phase 4
   inputs gate.
 
 **Reopener:** if a future inventory shows ≥ 3 shared-context loaders

package/docs/contracts/load-context-schema.md

@@ -33,10 +33,10 @@ load_context_eager:          # opt-in eager — auto-loaded on rule fire
 ---
 ```
 
-> **Logical names only.** The `.agent-src.uncompressed/` prefix is
+> **Logical names only.** The `.agent-src.uncondensed/` prefix is
 > rejected by the schema regex (`scripts/schemas/rule.schema.json`) and
-> by `scripts/lint_load_context.py`. The compress-time rewriter
-> (`scripts/compress.py::_rewrite_paths`) resolves logical names to the
+> by `scripts/lint_load_context.py`. The condense-time rewriter
+> (`scripts/condense.py::_rewrite_paths`) resolves logical names to the
 > deployment-correct relative path (e.g. `../contexts/<area>/<file>.md`
 > for a rule at `rules/<name>.md`). The full migration history lives
 > in the archived path-fixes roadmap under `agents/roadmaps/archive/`.
@@ -53,16 +53,16 @@ opt-in and budget-gated.
 ## Path rules
 
 - Logical names rooted at the source. The rewriter resolves them at
-  compress time.
+  condense time.
 - Paths MUST end in `.md`.
 - Allowed roots:
   - `contexts/<area>/<file>.md` — canonical logical name (resolves
-    against `.agent-src.uncompressed/`).
+    against `.agent-src.uncondensed/`).
   - `agents/settings/contexts/<file>.md` — project-local material (consumer
     repo only).
-  - `.agent-src/contexts/<file>.md` — compressed mirror; tolerated
+  - `.agent-src/contexts/<file>.md` — condensed mirror; tolerated
     defensively, not authored.
-- The `.agent-src.uncompressed/` prefix is **rejected** by the schema
+- The `.agent-src.uncondensed/` prefix is **rejected** by the schema
   regex and by `lint_load_context.py` with a remediation hint pointing
   at the canonical logical name.
 - A rule MAY reference contexts under either tree, but a package-shipped
@@ -144,10 +144,10 @@ definition it is not loaded into every turn.
 
 ## Public-vs-internal leak
 
-A rule shipped to consumers (`.agent-src.uncompressed/rules/`) may
+A rule shipped to consumers (`.agent-src.uncondensed/rules/`) may
 declare `load_context:` entries pointing at:
 
-- `.agent-src.uncompressed/contexts/` — public, OK.
+- `.agent-src.uncondensed/contexts/` — public, OK.
 - `agents/settings/contexts/` — package-internal, **lint warning** (the entry
   will not exist in consumer projects).
 
@@ -163,7 +163,7 @@ defined across both `load_context:` and `load_context_eager:` edges.
 
 ### Real consumer — `autonomous-execution`
 
-`.agent-src.uncompressed/rules/autonomous-execution.md` is the first
+`.agent-src.uncondensed/rules/autonomous-execution.md` is the first
 production rule to declare `load_context:`. Its frontmatter:
 
 ```yaml
@@ -173,9 +173,9 @@ description: "Deciding whether to ask the user or just act on a workflow step
 alwaysApply: false
 source: package
 load_context:
-  - .agent-src.uncompressed/contexts/execution/autonomy-detection.md
-  - .agent-src.uncompressed/contexts/execution/autonomy-mechanics.md
-  - .agent-src.uncompressed/contexts/execution/autonomy-examples.md
+  - .agent-src.uncondensed/contexts/execution/autonomy-detection.md
+  - .agent-src.uncondensed/contexts/execution/autonomy-mechanics.md
+  - .agent-src.uncondensed/contexts/execution/autonomy-examples.md
 ---
 ```
 

package/docs/contracts/local-analytics.md

@@ -1,3 +1,7 @@
+---
+stability: experimental
+---
+
 # Local Analytics Contract
 
 > **Status** · v0 / design · 2026-05-24. Phase 7 of the

package/docs/contracts/local-knowledge-ingestion.md

@@ -1,6 +1,6 @@
 ---
 stability: beta
-keep-beta-until: 2026-08-24
+keep-beta-until: 2026-08-23
 ---
 
 # Local knowledge ingestion contract

package/docs/contracts/mcp-cloud-scope.md

@@ -45,7 +45,7 @@ prose-only.
   `list_*`, `read_resource_body`) that touch the content blob only.
 - **What it never does:** execute Python scripts, shell out, spawn
   runtimes, touch consumer FS, write to R2, mutate consumer state,
-  call upstream LLM APIs, or read `.agent-src.uncompressed/`.
+  call upstream LLM APIs, or read `.agent-src.uncondensed/`.
 - **Owner code path:** `internal/workers/mcp/` (TypeScript, Cloudflare Worker).
   This contract is the normative spec.
 - **Auth model:** `public` (default) or `bearer-auth` (operator opt-in)
@@ -120,7 +120,7 @@ prose-only.
   shape `releases/v<X.Y.Z>-<sha>/` (immutable per release). The blob
   bundles `.agent-src/skills/<name>/SKILL.md`,
   `.agent-src/commands/**/*.md`, and `docs/guidelines/` (the same
-  projection the local kernel reads). Never reads `.agent-src.uncompressed/`.
+  projection the local kernel reads). Never reads `.agent-src.uncondensed/`.
 - **Identity surface:** `serverInfo.version` reads from a Worker-
   bundled constant, `_meta.packageVersion` reads from a
   `wrangler.toml` env var, `_meta.skillSetSignature` reads from a

package/docs/contracts/mcp-phase-1-scope.md

@@ -26,8 +26,8 @@ phase with its own design-call gate.
 - Transport: **stdio**. No SSE, no HTTP, no WebSocket.
 - MCP primitives: **`prompts/list`** + **`prompts/get`** — read-only.
 - Source data: **`.agent-src/skills/<name>/SKILL.md`** and
-  **`.agent-src/commands/**/*.md`** (compressed projections, never the
-  uncompressed source-of-truth tree).
+  **`.agent-src/commands/**/*.md`** (condensed projections, never the
+  uncondensed source-of-truth tree).
 - Loaded set (Phase 2): every well-formed skill + command under
   `.agent-src/` (Phase 1 hand-picked set retained as a smoke fixture
   in `prompts.py::PHASE_1_SKILLS`).
@@ -154,7 +154,7 @@ change invalidates it.
 
 ### F3 — Stdio Docker bundle
 
-`docker/mcp-server/Dockerfile` ships a stdio-only image. The contract:
+`internal/docker/mcp-server/Dockerfile` ships a stdio-only image. The contract:
 
 - **No HTTP / SSE listener** in the image. Stdio is the only wire.
 - The image embeds `scripts/mcp_server/`, the two tool dependencies

package/docs/contracts/measurement-baseline.md

@@ -5,7 +5,7 @@ stability: stable
 # Measurement baseline — contract
 
 > **Status:** locked 2026-05-16 · **Owner:** `step-4-measurement-and-benchmark.md`
-> · **Cited by:** every P2 enforcement roadmap (skill rationalization G0, north-star G1, compression default decision).
+> · **Cited by:** every P2 enforcement roadmap (skill rationalization G0, north-star G1, condensation default decision).
 
 Single source of truth for what `task bench` measures, what counts as
 drift, and what unblocks enforcement. Read this before pinning a number
@@ -36,7 +36,7 @@ timestamped, never overwritten, content-addressed by run.
 
 Total 26 prompts ≥ Acceptance Criteria floor of 25. Mid-window edits
 to either YAML restart the 60-day clock per
-[`compression-default-kill-criterion.md`](compression-default-kill-criterion.md) § 2.
+[`condensation-default-kill-criterion.md`](condensation-default-kill-criterion.md) § 2.
 
 ## What counts as drift
 
@@ -77,14 +77,14 @@ on the 30-report floor.
 Downstream consumers:
 
 - ``step-99-north-star-restructure.md` § Acceptance G1` — reads this exit code.
-- [`compression-default-kill-criterion.md` § 3](compression-default-kill-criterion.md) — reads the decision table after baseline closes.
+- [`condensation-default-kill-criterion.md` § 3](condensation-default-kill-criterion.md) — reads the decision table after baseline closes.
 - ``step-2-skill-inventory-rationalization.md` § G0` — usage-data soak floor.
 
 ## What the closeout writes
 
 On baseline closure, the step-4 closeout writes the numeric verdict to
 [`docs/parity/bench.json`](../parity/bench.json) — frozen snapshot with
-the 30+ reports averaged, drift verdict, and the compression-default
+the 30+ reports averaged, drift verdict, and the condensation-default
 decision per the kill-criterion table. That file is the artefact every
 P2 roadmap reads — not the live `internal/bench/reports/` directory.
 
@@ -98,5 +98,5 @@ P2 roadmap reads — not the live `internal/bench/reports/` directory.
 
 - [`benchmark-report-schema.md`](benchmark-report-schema.md) · per-report JSON schema
 - [`benchmark-corpus-spec.md`](benchmark-corpus-spec.md) · corpus YAML schema
-- [`compression-default-kill-criterion.md`](compression-default-kill-criterion.md) · decision table read by step-4 closeout
+- [`condensation-default-kill-criterion.md`](condensation-default-kill-criterion.md) · decision table read by step-4 closeout
 - `step-4-measurement-and-benchmark.md` · the owning roadmap

package/docs/contracts/mental-models.md

@@ -37,7 +37,7 @@ decisions. The expensive part is identifying which "principle" is
 actually load-bearing vs. inherited belief; the cheap part is the
 re-derivation. Use when an inherited approach feels stale and you
 suspect the real constraint moved.
-*Cited by:* `.agent-src.uncompressed/skills/improve-before-implement/`
+*Cited by:* `.agent-src.uncondensed/skills/improve-before-implement/`
 (challenges weak requirements before code is written).
 
 ### 2. Jobs-to-be-Done (JTBD)
@@ -47,7 +47,7 @@ the job is the situation × motivation × expected outcome, not the
 demographic. The unit of analysis is the **switch event** — what
 caused them to fire the previous solution. JTBD reframes feature
 requests as evidence of an unmet job, not feature gaps.
-*Cited by:* `.agent-src.uncompressed/skills/po-discovery/`
+*Cited by:* `.agent-src.uncondensed/skills/po-discovery/`
 (reframes fuzzy product asks via job-shape).
 
 ### 3. Pareto principle (80/20)
@@ -56,7 +56,7 @@ Roughly 80% of effects come from 20% of causes. The lift is in
 **identifying** the 20% — which user segment, which test failure,
 which N+1 query — not in re-stating the ratio. Anti-pattern: using
 80/20 as permission to ignore the long tail without measuring it.
-*Cited by:* `.agent-src.uncompressed/skills/performance-analysis/`
+*Cited by:* `.agent-src.uncondensed/skills/performance-analysis/`
 (N+1 detection prioritizes the 20% of queries causing 80% of latency).
 
 ### 4. Second-order thinking
@@ -66,7 +66,7 @@ what looks best now; second-order weighs the consequences of the
 consequences. Most "obvious" decisions die at second-order — the
 optimization that ships now creates the maintenance debt that kills
 velocity in 6 months.
-*Cited by:* `.agent-src.uncompressed/skills/adversarial-review/`
+*Cited by:* `.agent-src.uncondensed/skills/adversarial-review/`
 (stress-tests a plan by walking past the immediate verdict).
 
 ### 5. Opportunity cost
@@ -76,7 +76,7 @@ not pick**, not the dollar / time spent. A 2-week feature is not
 "2 weeks expensive" — it is the highest-value 2-week feature you
 chose not to ship instead. Naming the alternative makes the cost
 legible; pretending there isn't one is the failure mode.
-*Cited by:* `.agent-src.uncompressed/skills/rice-prioritization/`
+*Cited by:* `.agent-src.uncondensed/skills/rice-prioritization/`
 (scores compete for capacity; non-shipped items are the cost basis).
 
 ### 6. Theory of constraints
@@ -86,7 +86,7 @@ Improving any non-constraint resource is local optimization with
 zero system effect — usually negative, since it loads the actual
 constraint harder. The discipline: identify, exploit, subordinate,
 elevate, then find the next constraint.
-*Cited by:* `.agent-src.uncompressed/skills/funnel-analysis/`
+*Cited by:* `.agent-src.uncondensed/skills/funnel-analysis/`
 (identifies the single funnel stage that bounds conversion throughput).
 
 ### 7. MVP (Minimum Viable Product)
@@ -97,7 +97,7 @@ junior-grade product. The trap is shipping an MVP that cannot
 distinguish "user behavior validates the hypothesis" from "user
 behavior was driven by something else"; that is just a small
 product.
-*Cited by:* `.agent-src.uncompressed/skills/po-discovery/`
+*Cited by:* `.agent-src.uncondensed/skills/po-discovery/`
 (scopes a discovery slice that produces a learning, not a feature).
 
 ### 8. Build-Measure-Learn
@@ -107,7 +107,7 @@ to persevere or pivot. The skill is in **shortening the loop** —
 weeks beat months because months let teams rationalize a failed
 hypothesis. The loop is the artefact; any individual cycle is just
 one iteration.
-*Cited by:* `.agent-src.uncompressed/skills/test-driven-development/`
+*Cited by:* `.agent-src.uncondensed/skills/test-driven-development/`
 (the build-measure-learn loop applied to one function at a time).
 
 ### 9. Hypothesis-driven development
@@ -116,7 +116,7 @@ State the hypothesis as a falsifiable sentence with a metric and a
 threshold **before** writing the code. "If we do X, metric Y will
 move by Z." Without the threshold, the team will declare any
 movement victory. With it, the team learns from the misses.
-*Cited by:* `.agent-src.uncompressed/skills/project-analysis-hypothesis-driven/`
+*Cited by:* `.agent-src.uncondensed/skills/project-analysis-hypothesis-driven/`
 (competing hypotheses + validation loops + evidence-based conclusions).
 
 ### 10. Reversible vs. irreversible decisions
@@ -126,7 +126,7 @@ plus speed. The bias under uncertainty: shipping a reversible
 decision early is almost always cheaper than the meeting required
 to decide. The skill is recognizing irreversibility — usually data
 shape, public API surface, or hiring.
-*Cited by:* `.agent-src.uncompressed/skills/decision-record/`
+*Cited by:* `.agent-src.uncondensed/skills/decision-record/`
 (records the reversal-criteria so the irreversibility verdict is auditable).
 
 ### 11. DX as first-class concern
@@ -136,7 +136,7 @@ a polish task. Slow tests, fragile local setup, and surprising tool
 output compound — every developer pays the tax every day. Treat DX
 issues like user-facing bugs, with severity and SLA. The compounding
 math makes "fix it later" almost always wrong.
-*Cited by:* `.agent-src.uncompressed/skills/test-performance/`
+*Cited by:* `.agent-src.uncondensed/skills/test-performance/`
 (test-suite latency is a developer-facing metric, optimized as such).
 
 ### 12. Conway's Law
@@ -145,7 +145,7 @@ The systems an organization builds mirror its communication structure.
 Re-orgs propagate to architecture; architecture changes that fight the
 org chart fail. The lever is bidirectional: pick the architecture you
 want, then engineer the communication paths that produce it.
-*Cited by:* `.agent-src.uncompressed/skills/api-design/`
+*Cited by:* `.agent-src.uncondensed/skills/api-design/`
 (bounded-context choices follow team boundaries, not the other way around).
 
 ### 13. Occam's Razor
@@ -155,7 +155,7 @@ that introduces the fewest new entities. In debugging, the boring
 explanation (typo, off-by-one, stale cache) is usually correct. The
 trap is **assuming** instead of **falsifying** — Occam suggests
 order of investigation, not a verdict.
-*Cited by:* `.agent-src.uncompressed/skills/systematic-debugging/`
+*Cited by:* `.agent-src.uncondensed/skills/systematic-debugging/`
 (reproduce → isolate → hypothesize, simplest hypothesis first).
 
 ### 14. Meadows leverage points
@@ -165,7 +165,7 @@ system are paradigm shifts, then goals, then rules — far above
 parameter tweaks. Most "improvement" effort fights parameters
 (numbers, delays) at the bottom of the ranking. Climb the ranking
 before optimizing.
-*Cited by:* `.agent-src.uncompressed/skills/architecture-review-lens/`
+*Cited by:* `.agent-src.uncondensed/skills/architecture-review-lens/`
 (boundary / dependency-direction issues are higher-leverage than tweaks).
 
 ### 15. Signal vs. noise
@@ -175,7 +175,7 @@ A change of size N is meaningful only if N exceeds the noise floor
 for that metric × that horizon. The discipline: estimate the noise
 band first, then evaluate the change. Without the noise band, every
 movement looks like a trend.
-*Cited by:* `.agent-src.uncompressed/skills/funnel-analysis/`
+*Cited by:* `.agent-src.uncondensed/skills/funnel-analysis/`
 (stage-to-stage drop is read against the typical week-on-week noise).
 
 ### 16. Leading vs. lagging indicators
@@ -185,7 +185,7 @@ late. Leading indicators (activation events, repeat-use, support
 volume) are noisy and early. Operating on lagging alone means the
 team learns about problems after they cost money. Pair them: lagging
 is the score, leading is the steering wheel.
-*Cited by:* `.agent-src.uncompressed/skills/dashboard-design/`
+*Cited by:* `.agent-src.uncondensed/skills/dashboard-design/`
 (RED / USE / Golden Signals split leading from lagging explicitly).
 
 ### 17. Churn as health metric
@@ -195,7 +195,7 @@ delivers ongoing value. A flat retention tail means the product
 works for the survivors; a sliding tail means the underlying job is
 not getting done. Churn is upstream of CAC payback — fix it first,
 then scale.
-*Cited by:* `.agent-src.uncompressed/skills/funnel-analysis/`
+*Cited by:* `.agent-src.uncondensed/skills/funnel-analysis/`
 (retention bend distinguishes activation problems from product-fit).
 
 ### 18. Pull vs. push systems
@@ -205,7 +205,7 @@ start work when upstream capacity is free. Push optimizes individual
 utilization, pull optimizes flow. Most software teams claim pull and
 operate push (queues filling up, sprints overcommitted, WIP
 unmanaged). The fix is WIP limits, not motivation.
-*Cited by:* `.agent-src.uncompressed/skills/laravel-horizon/`
+*Cited by:* `.agent-src.uncondensed/skills/laravel-horizon/`
 (queue-balance strategies are pull-vs-push policy in concrete form).
 
 ### 19. Shift-left
@@ -215,7 +215,7 @@ lifecycle — to the developer's machine, the PR, the design — where
 the cost of fixing is order-of-magnitude lower. The trade-off:
 shift-left adds friction at the front; the math holds when defect
 escape rate drops faster than the friction cost.
-*Cited by:* `.agent-src.uncompressed/skills/threat-modeling/`
+*Cited by:* `.agent-src.uncondensed/skills/threat-modeling/`
 (threats enumerated before implementation, not after pen-test).
 
 ### 20. Latency vs. throughput
@@ -225,7 +225,7 @@ not implementation-detail. Batch processing trades latency for
 throughput; real-time pipelines trade throughput for latency. The
 mistake is optimizing without naming which one matters for the user
 job at hand.
-*Cited by:* `.agent-src.uncompressed/skills/database/`
+*Cited by:* `.agent-src.uncondensed/skills/database/`
 (index strategy and query batching surface the trade-off explicitly).
 
 ### 21. Trust boundaries
@@ -235,7 +235,7 @@ trusted: client → server, tenant A → tenant B, free tier → paid
 tier, public endpoint → internal service. Threats enter at boundary
 crossings; defense lives at the crossing, not deeper. Drawing the
 boundary correctly is half the security work.
-*Cited by:* `.agent-src.uncompressed/skills/threat-modeling/`
+*Cited by:* `.agent-src.uncondensed/skills/threat-modeling/`
 (produces trust boundaries + abuse cases mapped to files).
 
 ### 22. Defense in depth
@@ -245,7 +245,7 @@ rules, environment hardening, and instrumentation so a bypass at
 one layer is caught at the next. The trap is mistaking redundancy
 for security theatre — each layer must have an independent failure
 mode, not a copy of the previous one.
-*Cited by:* `.agent-src.uncompressed/skills/defense-in-depth/`
+*Cited by:* `.agent-src.uncondensed/skills/defense-in-depth/`
 (turns local fix into structural one across four guard layers).
 
 ### 23. Blast radius
@@ -255,7 +255,7 @@ consumer, queue worker, API client, migration, and test that the
 change touches. The radius is the work; the diff is the artefact.
 Underestimating radius is the failure mode that breaks production
 when the change "looked small".
-*Cited by:* `.agent-src.uncompressed/skills/blast-radius-analyzer/`
+*Cited by:* `.agent-src.uncondensed/skills/blast-radius-analyzer/`
 (file:line citation per dependency, BEFORE the edit).
 
 ### 24. Iron triangle (scope / time / quality)
@@ -264,7 +264,7 @@ Scope, time, and quality are coupled — fix two, the third moves.
 Pretending to fix all three is how teams ship at low quality and
 call it on-time. The honest move is naming which two are fixed and
 which one absorbs the variance, BEFORE the work starts.
-*Cited by:* `.agent-src.uncompressed/skills/refine-ticket/`
+*Cited by:* `.agent-src.uncondensed/skills/refine-ticket/`
 (AC sharpening forces explicit scope decisions before estimation).
 
 ### 25. Definition of Done
@@ -274,7 +274,7 @@ work — tests, docs, deployment, comms. Without one, every team
 member ships their personal threshold and disputes downstream. The
 discipline is making it visible, agreed, and consistently applied —
 not the specific items on it.
-*Cited by:* `.agent-src.uncompressed/skills/verify-completion-evidence/`
+*Cited by:* `.agent-src.uncondensed/skills/verify-completion-evidence/`
 (fresh evidence is required before any "done" claim).
 
 ### 26. Postmortem-driven learning
@@ -284,7 +284,7 @@ the team that hides the incident. Blameless postmortems separate
 contribution from blame, surface systemic causes, and produce
 mitigations that reduce future incident rate — not just the count
 of one-off fixes.
-*Cited by:* `.agent-src.uncompressed/skills/incident-commander/`
+*Cited by:* `.agent-src.uncondensed/skills/incident-commander/`
 (severity framing + comms cadence + post-mortem skeleton).
 
 ### 27. Tech debt as interest
@@ -294,7 +294,7 @@ payment (the ongoing tax on velocity). Carrying the debt is the
 right call when the principal is repayable and the interest is
 small; the failure is treating ongoing high-interest debt as a
 fixed cost. Track it like a balance sheet.
-*Cited by:* `.agent-src.uncompressed/skills/tech-debt-tracker/`
+*Cited by:* `.agent-src.uncondensed/skills/tech-debt-tracker/`
 (interest-vs-principal framing, prioritisation by carrying cost).
 
 ### 28. Mise en place
@@ -304,7 +304,7 @@ before the cooking step starts. Switching between prep and cook is
 where errors enter. In software, the analogue is staged commits,
 prepared test fixtures, and decisions locked before implementation.
 The discipline is the savings, not the metaphor.
-*Cited by:* `.agent-src.uncompressed/skills/existing-ui-audit/`
+*Cited by:* `.agent-src.uncondensed/skills/existing-ui-audit/`
 (inventory before any non-trivial UI edit, hard gate).
 
 ### 29. Premortem
@@ -314,7 +314,7 @@ risks the team already knows but has not voiced. The trick is the
 past-tense framing; "what could go wrong" surfaces less than "it
 failed, what happened". Cheap; high signal-to-noise; the residual
 output is a risk register the team actually defends.
-*Cited by:* `.agent-src.uncompressed/skills/risk-officer/`
+*Cited by:* `.agent-src.uncondensed/skills/risk-officer/`
 (blast-radius framing, mitigations, residual-risk verdict pre-commit).
 
 ### 30. Inversion
@@ -324,7 +324,7 @@ guarantee failure at X?" — then avoid that. Inversion exposes the
 non-obvious failure modes that direct planning misses, especially
 in security, ops, and people problems where the failure surface is
 larger than the success surface.
-*Cited by:* `.agent-src.uncompressed/skills/adversarial-review/`
+*Cited by:* `.agent-src.uncondensed/skills/adversarial-review/`
 (devil's-advocate stress-test poking holes in a plan).
 
 ## Adding or removing a model