@pasosdejesus/mpdj 0.3.2 → 0.3.3

package/SKILL.md

@@ -113,6 +113,10 @@ Reads files and copies content to clipboard.
 ./bin/m ai:readfile <file> [--head <N>] [--tail <N>] [--offset <N>] [--lines <N>] [--raw] [--no-clip]
 ```
 
+A range is `--offset <start> --lines <count>`: there is **no** `--lines 60-75`.
+Ask for the narrowest window that contains the block, so the agent receives the
+file's real line breaks (see rule 6 of `ai:edit`).
+
 ### ai:writefile
 
 Creates/replaces files from clipboard content.
@@ -123,14 +127,57 @@ Creates/replaces files from clipboard content.
 
 ### ai:edit
 
-Applies changes from clipboard (3 formats: full block, SEARCH/REPLACE, unified diff).
-
-> [WARN] Paths are relative to current directory. `___SEARCH___` must be exact copy including whitespace. Fuzzy matching (85%).
+Applies the changes a chat generated, from the clipboard or from a file.
 
 ```bash
-./bin/m ai:edit
+./bin/m ai:edit                       # clipboard
+./bin/m ai:edit --input answer.txt    # same thing, reproducible
+./bin/m ai:edit --dry-run             # show what would change, touch nothing
+```
+
+**The format that works** — one block per file, several blocks in one paste are
+fine:
+
+```text
+## Edit: resources/en/guide1.md
+___BUSCAR_20261030T1200___
+exact text taken from the file, indentation included
+___REMPLAZAR_20261030T1200___
+the new text
+___FIN_20261030T1200___
 ```
 
+Rules, in the order they usually bite:
+
+1. **The path is relative to where the command runs.** Run `ai:edit` from the
+   project root, or write the path as that directory sees it. When the file is
+   not there, `ai:edit` looks in the parent directories and prints a hint
+   (`[HINT]` / `[PISTA]`) with the path that does exist.
+2. **`___BUSCAR_` must be an exact copy of the file**, whitespace included. Get
+   it with `./bin/m ai:readfile <file>`; never rebuild it from memory. Matching
+   is fuzzy only above 85% and it says so.
+3. **The three markers must carry the same stamp.** The stamp itself is free
+   (`20261030`, `20261030T1200`, an ISO date); it only groups the markers.
+4. **Triple-backtick fences around the block are decoration**: optional, and
+   balanced if used. Never wrap the block in a second fence "so it looks
+   right": paste from `## Edit:` through `___FIN_` and nothing else.
+5. One block, one file. No prose inside the block, no partial blocks, no
+   `...` elisions: an elided block will not match.
+6. **Do not reflow the text.** Line breaks are part of `___BUSCAR_`: a markdown
+   paragraph is 4 long lines in the file and the chat is tempted to "wrap it
+   nicely" into 5 short ones. That drops the match to 90-95 % (fuzzy), applies
+   anyway, and rewrites the file's line breaks — a clean diff becomes noise. If
+   the wording is exact but the break is not, the block is still wrong.
+
+Other accepted formats: `## Archivo:` + the whole file inside a fence, or a
+unified diff.
+
+When the paste cannot be understood, `ai:edit` says **what is wrong** (the fence
+that is never closed, with its line number; a missing marker; stamps that
+differ; a missing header) instead of a generic "no changes found". When a block
+does not match the file, it copies the diagnostic to the clipboard for the
+agent to read.
+
 ### ai:grep
 
 Searches text in files (or clipboard if no `--path`).
@@ -151,8 +198,12 @@ File search by glob pattern.
 
 1. **Ask, don't order.**
 2. **Always use `ai:readfile` before suggesting changes.**
-3. **Use `ai:edit` with SEARCH/REPLACE** for specific changes.
+3. **Use `ai:edit` with SEARCH/REPLACE** for specific changes, in the exact shape
+   of `### ai:edit` above: `## Edit: <path>`, the three markers with the same
+   stamp, and an exact `___BUSCAR_` block.
 4. **NEVER assume content.** If replacement fails, `ai:edit` copies diagnostic to clipboard.
+5. **Give the whole block, never a fragment**: from `## Edit:` through
+   `___FIN_`, with no extra fence and no ellipses.
 
 ### Recommended workflow
 
@@ -208,7 +259,7 @@ If you have direct terminal and filesystem access (Crush, Cursor, etc.), **do no
                                                # --install, --build, --prepare (opt-in); --ready is an alias
 ./bin/m eng:verify [<motor>]                   # Guards the generated file against the sources and the intention
 ./bin/m eng:remove <motor> [--drop]            # Removes a vendored copy (--drop also drops its model tables)
-./bin/m eng:check [--runtime]                  # Schema vs DB + vendored drift + self-containment; --runtime probes LOAD_*
+./bin/m eng:check [--runtime] [--no-db]         # Schema vs DB + vendored drift + self-containment; --runtime probes LOAD_*
 ./bin/m eng:setup [<motor>] [--check]          # Generic: writes the motor migrations (db/migrations)
 ./bin/m eng:seed [<motor>] [--force] [--check] # Generic: loads db/seed.sql.gz; msip keeps its idempotent seed
 ./bin/m eng:msip:setup                         # Writes the `msip` base model migrations into the consumer
@@ -305,6 +356,8 @@ Full recipe: [doc/engines.md](../../doc/engines.md) §12; release ritual:
 - **OS:** OpenBSD/adJ (primary), Linux/Ubuntu (secondary, see `doc/ubuntu-compat.md`)
 - **pnpm:** v11 (pinned via `packageManager` and `pnpm-workspace.yaml`; v11 requires `allowBuilds` and ignores the `pnpm` field of `package.json`, see `doc/pnpm.md`)
 - **Build:** `pnpm run build`, **Tests:** `make test`
+- **Release:** `make release-check` (manifests: optional deps and peers use ranges) and
+  `make publish PKGS="..."`; ritual in [doc/version-bump.md](../../doc/version-bump.md)
 - Rollup WASM: `npm:@rollup/wasm-node` in `pnpm-workspace.yaml`
 - RPC: `NEXT_PUBLIC_RPC_URL` or `RPC_URL` in `.env`
 - PostgreSQL credentials for tests in `tests/test-project/apps/.env.test`
@@ -314,10 +367,16 @@ Full recipe: [doc/engines.md](../../doc/engines.md) §12; release ritual:
 ```bash
 ./bin/m contract:test [--fork-url <url>] [--fork-block <N>] [--verbose]
 ./bin/m contract:compile [--solc <path>|--solc-js <soljson>|--solc-version <v>]
+./bin/m contract:typechain [--target ethers-v6] [--out-dir typechain-types]
+./bin/m contract:node [--host 127.0.0.1] [--port 8545]
+./bin/m contract:deploy --contract <Name> --network <net> [--rpc <url>] [--args '[...]']
+./bin/m contract:verify --contract <Name> --network <net> [--api-url <url>] [--dry-run]
+./bin/m contract:clean
 ./bin/m contract:checkdeploy [--json|--strict|--root <dir>|--network <name>]
 ```
 Runs Hardhat-compatible tests on OpenBSD/adJ using `@ethereumjs/vm` (EDR replacement);
-`contract:compile` compiles without Hardhat and `contract:checkdeploy` validates deployment addresses.
+compiles, generates types, runs a local node, deploys/verifies with viem and cleans without Hardhat;
+`contract:checkdeploy` validates deployment addresses.
 See [doc/edr.md](../../doc/edr.md) and [issue #7](https://gitlab.com/pasosdeJesus/m/-/issues/7).
 
 ### E2E Test Commands

package/dist/SKILL.md

@@ -113,6 +113,10 @@ Reads files and copies content to clipboard.
 ./bin/m ai:readfile <file> [--head <N>] [--tail <N>] [--offset <N>] [--lines <N>] [--raw] [--no-clip]
 ```
 
+A range is `--offset <start> --lines <count>`: there is **no** `--lines 60-75`.
+Ask for the narrowest window that contains the block, so the agent receives the
+file's real line breaks (see rule 6 of `ai:edit`).
+
 ### ai:writefile
 
 Creates/replaces files from clipboard content.
@@ -123,14 +127,57 @@ Creates/replaces files from clipboard content.
 
 ### ai:edit
 
-Applies changes from clipboard (3 formats: full block, SEARCH/REPLACE, unified diff).
-
-> [WARN] Paths are relative to current directory. `___SEARCH___` must be exact copy including whitespace. Fuzzy matching (85%).
+Applies the changes a chat generated, from the clipboard or from a file.
 
 ```bash
-./bin/m ai:edit
+./bin/m ai:edit                       # clipboard
+./bin/m ai:edit --input answer.txt    # same thing, reproducible
+./bin/m ai:edit --dry-run             # show what would change, touch nothing
+```
+
+**The format that works** — one block per file, several blocks in one paste are
+fine:
+
+```text
+## Edit: resources/en/guide1.md
+___BUSCAR_20261030T1200___
+exact text taken from the file, indentation included
+___REMPLAZAR_20261030T1200___
+the new text
+___FIN_20261030T1200___
 ```
 
+Rules, in the order they usually bite:
+
+1. **The path is relative to where the command runs.** Run `ai:edit` from the
+   project root, or write the path as that directory sees it. When the file is
+   not there, `ai:edit` looks in the parent directories and prints a hint
+   (`[HINT]` / `[PISTA]`) with the path that does exist.
+2. **`___BUSCAR_` must be an exact copy of the file**, whitespace included. Get
+   it with `./bin/m ai:readfile <file>`; never rebuild it from memory. Matching
+   is fuzzy only above 85% and it says so.
+3. **The three markers must carry the same stamp.** The stamp itself is free
+   (`20261030`, `20261030T1200`, an ISO date); it only groups the markers.
+4. **Triple-backtick fences around the block are decoration**: optional, and
+   balanced if used. Never wrap the block in a second fence "so it looks
+   right": paste from `## Edit:` through `___FIN_` and nothing else.
+5. One block, one file. No prose inside the block, no partial blocks, no
+   `...` elisions: an elided block will not match.
+6. **Do not reflow the text.** Line breaks are part of `___BUSCAR_`: a markdown
+   paragraph is 4 long lines in the file and the chat is tempted to "wrap it
+   nicely" into 5 short ones. That drops the match to 90-95 % (fuzzy), applies
+   anyway, and rewrites the file's line breaks — a clean diff becomes noise. If
+   the wording is exact but the break is not, the block is still wrong.
+
+Other accepted formats: `## Archivo:` + the whole file inside a fence, or a
+unified diff.
+
+When the paste cannot be understood, `ai:edit` says **what is wrong** (the fence
+that is never closed, with its line number; a missing marker; stamps that
+differ; a missing header) instead of a generic "no changes found". When a block
+does not match the file, it copies the diagnostic to the clipboard for the
+agent to read.
+
 ### ai:grep
 
 Searches text in files (or clipboard if no `--path`).
@@ -151,8 +198,12 @@ File search by glob pattern.
 
 1. **Ask, don't order.**
 2. **Always use `ai:readfile` before suggesting changes.**
-3. **Use `ai:edit` with SEARCH/REPLACE** for specific changes.
+3. **Use `ai:edit` with SEARCH/REPLACE** for specific changes, in the exact shape
+   of `### ai:edit` above: `## Edit: <path>`, the three markers with the same
+   stamp, and an exact `___BUSCAR_` block.
 4. **NEVER assume content.** If replacement fails, `ai:edit` copies diagnostic to clipboard.
+5. **Give the whole block, never a fragment**: from `## Edit:` through
+   `___FIN_`, with no extra fence and no ellipses.
 
 ### Recommended workflow
 
@@ -208,7 +259,7 @@ If you have direct terminal and filesystem access (Crush, Cursor, etc.), **do no
                                                # --install, --build, --prepare (opt-in); --ready is an alias
 ./bin/m eng:verify [<motor>]                   # Guards the generated file against the sources and the intention
 ./bin/m eng:remove <motor> [--drop]            # Removes a vendored copy (--drop also drops its model tables)
-./bin/m eng:check [--runtime]                  # Schema vs DB + vendored drift + self-containment; --runtime probes LOAD_*
+./bin/m eng:check [--runtime] [--no-db]         # Schema vs DB + vendored drift + self-containment; --runtime probes LOAD_*
 ./bin/m eng:setup [<motor>] [--check]          # Generic: writes the motor migrations (db/migrations)
 ./bin/m eng:seed [<motor>] [--force] [--check] # Generic: loads db/seed.sql.gz; msip keeps its idempotent seed
 ./bin/m eng:msip:setup                         # Writes the `msip` base model migrations into the consumer
@@ -305,6 +356,8 @@ Full recipe: [doc/engines.md](../../doc/engines.md) §12; release ritual:
 - **OS:** OpenBSD/adJ (primary), Linux/Ubuntu (secondary, see `doc/ubuntu-compat.md`)
 - **pnpm:** v11 (pinned via `packageManager` and `pnpm-workspace.yaml`; v11 requires `allowBuilds` and ignores the `pnpm` field of `package.json`, see `doc/pnpm.md`)
 - **Build:** `pnpm run build`, **Tests:** `make test`
+- **Release:** `make release-check` (manifests: optional deps and peers use ranges) and
+  `make publish PKGS="..."`; ritual in [doc/version-bump.md](../../doc/version-bump.md)
 - Rollup WASM: `npm:@rollup/wasm-node` in `pnpm-workspace.yaml`
 - RPC: `NEXT_PUBLIC_RPC_URL` or `RPC_URL` in `.env`
 - PostgreSQL credentials for tests in `tests/test-project/apps/.env.test`
@@ -314,10 +367,16 @@ Full recipe: [doc/engines.md](../../doc/engines.md) §12; release ritual:
 ```bash
 ./bin/m contract:test [--fork-url <url>] [--fork-block <N>] [--verbose]
 ./bin/m contract:compile [--solc <path>|--solc-js <soljson>|--solc-version <v>]
+./bin/m contract:typechain [--target ethers-v6] [--out-dir typechain-types]
+./bin/m contract:node [--host 127.0.0.1] [--port 8545]
+./bin/m contract:deploy --contract <Name> --network <net> [--rpc <url>] [--args '[...]']
+./bin/m contract:verify --contract <Name> --network <net> [--api-url <url>] [--dry-run]
+./bin/m contract:clean
 ./bin/m contract:checkdeploy [--json|--strict|--root <dir>|--network <name>]
 ```
 Runs Hardhat-compatible tests on OpenBSD/adJ using `@ethereumjs/vm` (EDR replacement);
-`contract:compile` compiles without Hardhat and `contract:checkdeploy` validates deployment addresses.
+compiles, generates types, runs a local node, deploys/verifies with viem and cleans without Hardhat;
+`contract:checkdeploy` validates deployment addresses.
 See [doc/edr.md](../../doc/edr.md) and [issue #7](https://gitlab.com/pasosdeJesus/m/-/issues/7).
 
 ### E2E Test Commands

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@pasosdejesus/mpdj",
-  "version": "0.3.2",
+  "version": "0.3.3",
   "description": "Extensión Pasos de Jesús para m CLI — marco cristiano, credenciales, requerimientos",
   "author": "",
   "license": "ISC",
@@ -64,7 +64,7 @@
     "REQ"
   ],
   "peerDependencies": {
-    "@pasosdejesus/m": "^0.37.0",
+    "@pasosdejesus/m": "^0.37.9",
     "next": ">=15.0.0",
     "react": ">=18.0.0",
     "vitest": ">=2.1.9"