Skip to content

Find code by asking what it does. A CLI for coding agents that uses Jev to discover relevant files and source context.

skills(jevgrep): clarify query phrasing with behavioral examples

Harshavarma026a3b0919ee87
diff --git a/.agents/skills/jevgrep/SKILL.md b/.agents/skills/jevgrep/SKILL.md
index 1e8014c..f1fa9ee 100644
--- a/.agents/skills/jevgrep/SKILL.md
+++ b/.agents/skills/jevgrep/SKILL.md
@@ -26,6 +26,22 @@ jg "How are telemetry events recorded and sent?" .
Pass a natural-language question and an optional search root. The root defaults
to the current directory; a narrower folder limits the search to that subtree.
+
+### Phrasing Queries
+
+Frame questions around **behavior, lifecycle, and system interactions** (how, where,
+why, or what triggers a behavior). Jevgrep evaluates semantic relevance to find entry
+points across unfamiliar files:
+
+- **Effective queries (behavioral & explanatory):**
+ - `"Where is authentication checked before a request reaches a handler?"`
+ - `"How are database connections created, pooled, and closed?"`
+ - `"Which tests cover retry behavior when a request times out?"`
+- **Avoid keyword or symbol lookups:**
+ - Do not pass isolated keywords or symbol names (e.g. `"auth"` or `"AuthHandler"`).
+ Use `grep`, `ripgrep`, or code definitions instead.
+ - Do not ask for file listings or paths; use `jg files` or filesystem search.
+
When the root is large or its scope is unclear, run `jg files [root]` before
searching. It reports file/byte totals and top-level directory groups without
credentials or provider requests; it does not read source content. Counts are an
diff --git a/skills/jevgrep/SKILL.md b/skills/jevgrep/SKILL.md
index 1e8014c..f1fa9ee 100644
--- a/skills/jevgrep/SKILL.md
+++ b/skills/jevgrep/SKILL.md
@@ -26,6 +26,22 @@ jg "How are telemetry events recorded and sent?" .
Pass a natural-language question and an optional search root. The root defaults
to the current directory; a narrower folder limits the search to that subtree.
+
+### Phrasing Queries
+
+Frame questions around **behavior, lifecycle, and system interactions** (how, where,
+why, or what triggers a behavior). Jevgrep evaluates semantic relevance to find entry
+points across unfamiliar files:
+
+- **Effective queries (behavioral & explanatory):**
+ - `"Where is authentication checked before a request reaches a handler?"`
+ - `"How are database connections created, pooled, and closed?"`
+ - `"Which tests cover retry behavior when a request times out?"`
+- **Avoid keyword or symbol lookups:**
+ - Do not pass isolated keywords or symbol names (e.g. `"auth"` or `"AuthHandler"`).
+ Use `grep`, `ripgrep`, or code definitions instead.
+ - Do not ask for file listings or paths; use `jg files` or filesystem search.
+
When the root is large or its scope is unclear, run `jg files [root]` before
searching. It reports file/byte totals and top-level directory groups without
credentials or provider requests; it does not read source content. Counts are an